447JMed Genet 1997;34:447-452

The incidence of deafness is non-randomlydistributed among families segregating forWaardenburg syndrome type 1 (WS 1)

Robert Morell, Thomas B Friedman, James H Asher Jr, Leonard G Robbins

AbstractWaardenburg syndrome (WS) is caused byautosomal dominant mutations, and ischaracterised by pigmentary anomaliesand various defects ofneural crest derivedtissues. It accounts for over 2% ofcongeni-tal deafness. WS shows high variability inexpressivity within families and differ-ences in penetrance of clinical traitsbetween families. While mutations in thegene PAX3 seem to be responsible formost, if not all, WS type 1, it is still notclear what accounts for the reducedpenetrance of deafness. Stochastic eventsduring development may be the factorsthat determine whether a person with a

PAX3 mutation will be congenitally deafor not. Alternatively, genetic backgroundor non-random environmental factors or

both may be significant. We compared thelikelihoods for deafness in affected sub-jects from 24 families with reported PAX3mutations, and in seven of the familiesoriginally described by Waardenburg. Wefound evidence that stochastic variationalone does not explain the differences inpenetrances of deafness among WS fami-lies. Our analyses suggest that geneticbackground in combination with certainPAX3 alleles may be important factors inthe aetiology of deafness in WS.(7 Med Genet 1997;34:447-452)

Keywords: Waardenburg syndrome; WS1; deaffiess;PAX3

Genetics Program andDepartment ofZoology, MichiganState University, EastLansing, Michigan48824-1115, USAR Morell*T B Friedman*J H Asher JrL G Robbins

*Present address: Laboratoryof Molecular Genetics,National Institute onDeafness and OtherCommunication Disorders,National Institutes of Health,5 Research Court, Rockville,MD 20850, USA.

Correspondence to:Dr Morell.

Received 6 August 1996Revised version accepted forpublication 10 December1996

Waardenburg syndrome (WS) is caused byautosomal dominant mutations and is charac-terised by deafness, pigmentary anomalies, andvarious defects ofneural crest derived tissues.' '

It accounts for over 2% of the congenitally deafpopulation.3 At least four types (WS 1, 2, 3, and4) are recognised on the basis of clinicalattributes.2' For Waardenburg syndrome type1 (WS 1), telecanthus, or dystopia canthorummedialis, is the defining characteristic and has a

penetrance of greater than 90%. All of theclinical features, however, have incompletepenetrance and variable expressivity, even

within families segregating the same mutation.Waardenburg' originally estimated the pen-

etrance for deafness to be 20%, but more

recent estimates are 36% to 58% for type and57% to 74% for type 2.7 Large multigenerationWS families with penetrance greater than75% have also been reported.`'0 Estimates forthe penetrances of other clinical features of thesyndrome are: broad nasal root (78%), white

forelock (17%), heterochromia iridum (25%),and synophrys (45%).'WS1 maps to chromosome 2q351" where

mutations in the PAX3 gene have been reportedfor many type 1 and some type 3 patients.'0 12-20PAX3 encodes a transcription factor containingtwo DNA binding motifs: a paired domain anda paired type homeodomain." It is expressed indeveloping neural crest cells and in thebrain.2' 22 The PAX3 mutations described inWS or WS-like families include amino acidreplacements in both DNA binding domains aswell as deletions and insertions resulting inframeshifts.

Until recently, linkage analyses suggestedthat all WS 1 mutations cosegregate with PAX3markers, even in families for which no PAX3mutation has been identified." The descriptionof a family in which WS 1 is not linked to PAX3calls this assertion into question, although thisfamily is atypical ofWS 1 in also having cleft lipand Hirschsprung's disease.23There is no obvious correlation between the

nature of the mutations and the resultingphenotypes." This lack of correlation, coupledwith the variability of expression within fami-lies, suggests that stochastic events might play amajor role in the manifestation of deafness insubjects who are heterozygous for mutant alle-les. This hypothesis is supported by observa-tions in the mouse model for WS, Splotch,where the expression of coat colour patches,neural crest cell migration, and spinal ganglionvolumes show considerable variability even inhighly inbred laboratory strains.'425 It is alsoreasonable to consider the possibility thatmodifier genes, perhaps under the transcrip-tional regulation of PAX3, are responsible fordeafness. Surprisingly, the Splotch mouse alsoprovides supporting evidence for this hypoth-esis, in that Asher et al'6 have shown that genessegregating in an interspecific cross modifycraniofacial morphology in the presence of aPax-3 mutation. Wlhen comparing the merits ofthese two hypotheses, it would be useful to knowwhether the differences in the proportion of deafpeople between families is consistent withrandom fluctuation around a mean proportion,or whether deafness tends to cluster in certainfamilies. Clustering, if it occurs, might bebecause of the differential effects of eithergenetic background or environmental factors.We analysed the variation in penetrances for

deafness in 24 published families segregatingPAX3 mutations (table 1). We also analysedseven of the 14 families originally described byWaardenburg' (table 2). From the set of familiesdescribed by Waardenburg,' we excluded those

on Septem

ber 10, 2020 by guest. Protected by copyright.

http://jmg.bm

j.com/

J Med G

enet: first published as 10.1136/jmg.34.6.447 on 1 June 1997. D

ownloaded from

Morell, Friedman, Asher, Robbins

Table I PAX3 mutations and their reported penetrances of deafness. For each mutantallele the type and location of the mutation is listed as well as the numbers of heterozygotesthat are deafand hearing

Heterozygotes

Family Mutation Exon Deaf Hearing References

WS.055 F45L 2 3 4 32MSU17 N47L 2 3 0 33 34BU26 P50L 2 18 5 8 14BU35 R56L 2 3 0 14 35WS.15 G81A 2 4 0 920WS.009 G99D 2 2 1 32MSU3 100 ins 1 2 2 0 16WS.05 185 del 18 2 6 0 9 19UGM2 266 del 14 2 8 3 10WS.06 288 del 1 2 1 6 9 20BU53 297 del 28 2 3 4 36WS.084 451 ins 1 3 1 1 32WS.003 452 del 2 4 2 1 32WS.l1 556del2 4 6 2 920BU9 R223X 5 1 2 36WS.001 Q254X 5 5 2 9 37Hol, 1995 598 del 5 5 1 6 38NIH3 V265F 6 5 3 15WS.10 R271C 6 2 5 9MSU5 R271C 6 3 8 1831NIH8 R271G 6 5 1 15MSU7 W274X 6 3 1 18MSU9 874 ins 1 6 3 0 18 39BU25 1185 ins 3 8 8 8 8 13

Table 2 Penetrance of deafness among the WSI familiesreported by Waardenburg.' Families segregating for WStype 2 and those with probable new mutations are excluded.Asterisks indicate that two probands were reportedfor thefamily, thus, for likelihood calculations, the penetrance ofdeafness was corrected assuming a multiple incomplete biasof ascertainment

Affected subjects (WSI)

Family Deaf Hearing

5* 4 196* 3 17 1 29 1 210 1 1511* 8 3212 6 5

that were WS2, and those with only a singleaffected subject born to clinically normalparents. We used the log likelihood ratio test or

G testl7-29 to compare penetrances for deafnessamong WS families. Our analyses suggest thatthe aetiology of deafness is dependent on geneticor environmental background or both, and can-

not be attributed to stochastic events alone.

MethodsWe reviewed published reports describingPAX3 mutations. Data were included in theanalyses only from pedigrees where familyrelationships and sufficient clinical informationwere provided in either the mutation report or

a previous publication. Only affected people,that is, those shown to be heterozygous for a

PAX3 mutation or who had dystopia cantho-rum in the case of families described byWaardenburg,' were included. A person was

classified as either deaf or hearing based uponthe published designation. Audiologicalexamination information is not presented in

any of the cited references. In the pedigreesdescribed by Waardenburg,' the mutationshave neither been characterised nor mapped.The estimation of penetrance in these pedi-grees was corrected for ascertainment bias

under the assumption of multiple incompleteascertainment. 1Data on the hearing status of affected mem-

bers were entered into a spreadsheet with eachrow (i) being a family or a sibship, and eachcolumn 0) containing the number of deaf orhearing WS1 people for each row. The i x jtable was then smoothed to correct for numer-ous small and zero cell entries using themethod of pseudo-Bayes estimates recom-mended by Bishop et al.28 This accomplishesfor a large table what Yates correction does fora 2 x 2 table. However, instead of adding anarbitrary amount to each cell, the values areadjusted in a way more commensurate to thevariations in size of the cell entries. For eachcell xij, total sample size N, a weighting factor k,and an estimated value of the cell X>, the valuefor each cell in the corrected table is given by:

NXl + Xij k Ri..

where

N2- L'72;1i=ij N 2K)

For each table, the estimate for kiX used is theexpected proportion of affected subjects as-suming homogeneity of penetrance in thetable, or:

"IN and k i2 =E/N

Thus the entire table was biased towardshomogeneity by correcting each cell towardsthe overall column total (=1 or 2), that is,towards the overall probability of deafness andhearing respectively.The tables were then used for a series of

likelihood ratio tests, tests that are powerfuland economical with limited data. For exam-ple, to test whether the penetrance for deafnessis essentially the same in all families (hypoth-esis 1 or H1) or whether it is different for eachfamily (H2), we can construct two likelihoodfunctions:

L(H1) N()a(J_PN)b N

n

L(H2) =f P) ( pb

where for each family i, ai is the number ofaffected subjects who are deaf, bi is the numberof affected subjects with normal hearing, pi isthe maximum likelihood estimate for pen-etrance of deafness in that family, n is thenumber of families in the data set, andN refersto the entire pooled data set. For any set ofaffected subjects, pi is the number of deaf sub-jects divided by the total number of subjectswith the WS 1 mutation in a given family. Inother words, the ln likelihoods are calculated

448

on Septem

ber 10, 2020 by guest. Protected by copyright.

http://jmg.bm

j.com/

J Med G

enet: first published as 10.1136/jmg.34.6.447 on 1 June 1997. D

ownloaded from

Deafness inWS4

Table 3 Likelihoods of deafness and G tests of hypotheses. ln likelihoods were calculatedafter ascertainment bias correction and pseudo-Bayesian smoothing of the data tables asdescribed in the text

Hypothesis lnL No ofparameters

(A) For pedigrees with PAX3 mutationsHi Penetrance of deafness same for all PAX3 mutation

pedigrees -109 1H2 Penetrance of deafness different for each PAX3 mutation

pedigree -83.8 24H3 Penetrance of deafness different for each sibship of all

PAX3 mutation pedigrees -39.9 94Comparison G2 p (d])HlvH3 196 <1x10-8(76)HIv H2 50.5 0.0008 (23)H2 v H3 87.86 0.07 (70)

(B) For WSI pedigrees described by Waardenburg'H1' Penetrance of deafness same for all Waardenburg pedigrees -76.1 1H2' Penetrance of deafness different for each Waardenburg

pedigree -66.7 7H3' Penetrance of deafness different for each sibship of all

Waardenburg pedigrees -21.0 45Comparison G2 p (df)HI'v H3' 110 <1X10-8 (44)Hl'v H2' 18.8 0.004(6)H2'v H3' 91.4 <3x10-6(38)

(C) Comparison of the two setsH4 Penetrance same for the two data sets -216 1H5 Penetrance different between the PAX3 mutation and

Waardenburg' data sets -185 2Comparison G2 p (dl)H4 v H5 61.8 <1x10-8 (1)

for the summed data and for each row of thespreadsheet.The ln likelihoods of the hypotheses were

compared by the G test27 using the metric: G2= 2 x (lnL(H1) - lnL(H2)).G2 is distributed approximately as X2, So the

p value was calculated using the CHIPROBprogram.30 The number of the degrees of free-dom is the absolute value of the difference innumbers of parameters estimated in the twohypotheses being compared.28 For example, inHI one parameter is estimated: the averagepenetrance of deafness in the entire PAX3 dataset. In H2 the penetrance in each family isseparately estimated, so there are 24 para-meters. Thus when comparing Hi to H2 thereare 23 degrees of freedom. Similarly, in H3 thepenetrance of deafness in each sibship is a

separately estimated parameter.

Results and discussionThe first three columns of table 1 summarisemany of the PAX3 mutations published to date(with references provided in column 6).Reported mutations for which there is insuffi-cient published clinical information or no

pedigree structure information are not in-cluded in the table. The number of deaf andhearing subjects with PAX3 mutant alleles are

given in columns 4 and 5. The type ofmutationis listed in column 2, and it is notable that allbut one of the reported alleles are unique toeach family. The exception is the R27 1Cmutation in WS.10 and MSU5. WS.I0 is a

British family first reported by Foy et al,9 whileMSU5 is an American family reported byAsher et al.3" These two families do not appear

to be related as far as five generations back, buta haplotype analysis has not been done todetermine whether the alleles are likely to havethe same origin. The penetrance for deafness inthe two families is similar: there are two deaf

out of seven affected in WS. 10 and three deafout of 11 affected in MSU5. This contrastswith the penetrance for deafness (5/6) inNIH8, which segregates for a different mis-sense mutation at the same position (R271G).

Table 2 summarises the hearing status ofaffected subjects in the seven WS1 familiesdescribed by Waardenburg.'The question is whether the varying ratios of

deaf to hearing among the affected subjects ineach family reflects stochastic variation orwhether the probability of being deaf differsfrom family to family. The natural log likeli-hoods (lnL) for the penetrance of deafness ineach family in the PAX3 data set are shown intable 3A. The G test between HI and H2 givesa value of 50.5. Because 24 parameters areestimated in H2 (one for each family), and oneparameter is estimated in HI, there are 23degrees of freedom, resulting in a p value of0.0008.The heterogeneity among these 24 families

indicated by this test could be caused bygenetic or environmental effects, but it alsocould be caused by heterogeneity in ascertain-ment or in the diagnostic criteria used for"deafness". However, the diagnoses of deafnessare probably consistent because all but one ofthe reports38 are from laboratories participatingin a consortium with agreed criteria for clinicalassignment.40The families described by Waardenburg'

have the virtue of being ascertained anddiagnosed by a single clinician, with probandsidentified, thus reducing the potential forheterogeneity resulting from differing ascer-tainment or diagnostic criteria. The likelihoodsfor homogeneity among these families (Hi')and for heterogeneity (H2') are calculated withascertainment bias correction and the result ofa G test is reported in table 3B. Once again,significant heterogeneity in penetrance amongthe pedigrees is seen (p=0.0004).The heterogeneity in both data sets could be

ascribed to the effects ofunique alleles for eachfamily, differing environmental influences, ordiffering genetic backgrounds. The influenceof these factors needs not be evenly distributedamong sibships within pedigrees. Generally,one would expect that the effects caused by agiven mutant allele would affect all sibships ina pedigree evenly, environmental factors prob-ably less so, while background genetic factorswould be the most specific to sibships becauseof the unique genetic contributions of the par-ents who marry into the family. It is logical,then, to evaluate the hypothesis that thepenetrance of deafness is different for each sib-ship (hypotheses designated H3 for the PAX3data set and H3' for the Waardenburg data set)and compare this with the other hypotheses.

Table 3 summarises the lnLs for eachhypothesis and gives the G test results of com-parisons. Not surprisingly, comparisons of H1to H3 and HI' to H3' yield significant G statis-tics, since the heterogeneity already detectedamong pedigrees is implicit among the sibshipsof the pedigrees. The more telling comparisonsare H2 to H3 and H2' to H3', which test forheterogeneity of the penetrance of deafness in

449

on Septem

ber 10, 2020 by guest. Protected by copyright.

http://jmg.bm

j.com/

J Med G

enet: first published as 10.1136/jmg.34.6.447 on 1 June 1997. D

ownloaded from

Morell, Friedman, Asher, Robbins

Table 4 Analysis of variation ofpenetrance with respect to mutation type and location

Hypothesis InL No ofparameters

(A) Effects of mutation site on penetrance of deafniessHI Penetrance of deafness same regardless of mutation -107.8 1

location by exonH2 Penetrance of deafness different depending on exon in -104.0 6

which mutation occursConmparison G2 p (df)H1 vH2 7.62 0.178 (5)

(B) Effects of mutation type (missense or nonsenselframeshift) on penetrance of deafnessHI Penetrance of deafiess same regardless of type -107.8 1H2 Penetrance of deafiess different for missense and for -107.5 2

nonsense/frameshift mutationsConmparisol G'2 p (df)H1 vH2 0.6 0.44 (1)

sibships over and above the heterogeneityalready present among pedigrees. In otherwords, is there significant clustering of deafnesswithin the sibships of the pedigrees? Thiswould be expected if environmental or geneticbackground differences contributed to theprobability of being deaf, but would not beexpected if the nature of the mutant alleleswere the major determinant. In the PAX3mutation data set, no significant evidence forclustering of deaf within sibships was seen(p=0.07), but there was significant evidence inthe Waardenburg data set (p<3x10-6).Why is there a significant result for one data

set but not the other? Are the data setsessentially the same, or do they represent twodifferent "Waardenburg syndrome popula-tions" with respect to the frequency ofdeafness? Again, a G test can be performed bywriting likelihood statements for the probabil-ity of deafness over the two sets. The lnlikelihoods for the set comprised of all familiespooled together (H4) is compared with thesum of lnLs for the PAX3 mutation data set

Table 5 Tests for heterogeneity among pedigrees segregating missense mutations

Hypothesis lnL No ofparameters

(A) Tests for heterogeneity by anzino acid replacenment, site of replacement, and genetic backgrotundHI Penetrance of deafness same for all amino acid

replacements -50.2 1H2 Penetrance of deafness different for each amino acid

replacement (note: data for MSU5 and WS. 10 are

combined) -39.4 9H3 Penetrance of deafness same among amino acid

replacements in the same DNA binding domain, butdifferent for the two domains -47.2 2

H4 Penetrance of deafness different for each sibship -19.5 48Conmparison G p (dl)HlvH2 21.5 0.006(8)HlvH3 6.0 0.014(1)HIv H4 62.0 0.070 (47)H2 v H3 15.6 0.029 (7)H2 v H4 39.8 0.434 (39)H3vH4 55.4 0.161 (46)

(B) Pedigrees segregating anino acid replacenmenzts in the paired domainlHI Penetrance of deafness same for all alleles -24.1 1

H2 Penetrance of deafness different for each allele -20.1 6

H3 Penetrance of deafness different in each sibship -10.6 29

Comparison G2 p (dD)HI v H3 27.2 0.507 (28)HlvH2 8.2 0.146(5)H2 v H3 19.1 0.696 (23)

(C) Pedigrees segregating anzino acid replacenments in the honmeodonmai'lHI Penetrance of deafness same for all alleles -23.5 1

H2 Penetrance of deafness different for each allele -20.5 3

H3 Penetrance of deafness different in each sibship -11.1 19

Comparison G p (df)HlvH3 24.7 0.132(18)Hlv H2 6.0 0.050 (2)H2 vH3 18.8 0.173(14)

and the Waardenburg data set (lnL(H 1) +lnL(H 1') = lnL(H5)). The results confirm thatthe two data sets are significantly different(table 3C). This is intuitively obvious whencomparing the overall estimates of penetrancefor deafness in the two sets. The penetrance ofdeafness in the PAX3 mutation data set is 61%,and is best estimated as 18% for the Waarden-burg data set (corrected for multiple incom-plete ascertainment). It should be noted thatthe differences in penetrance between the twosets might contribute to a difference in thepower of the G test to detect heterogeneityamong sibships. Failure to find sibship hetero-geneity in the PAX3 set might also be the resultof differences in the number of sibships andaverage size of sibships. A larger number ofsibships would result in greater power to detectheterogeneity, but a lower average sibship sizewould decrease the power of the test. In thePAX3 set there are 77 sibships with average size1.66; in the Waardenburg set there are 45 sib-ships with an average size of 2.22.Does the penetrance of deafness depend on

type and location of mutation in PAX3? Table4 summarises penetrance of deafness whenmutations are categorised by location and thenby type. G tests show no significant effects onpenetrance of deafness when mutations arecategorised by exon location (table 4A). Norare significant differences observed if themutations are categorised by type, either asmissense or nonsense/frameshift (which in-cludes frameshift mutations and in frame dele-tions, as well as substitutions creating termina-tion codons) (table 4B). The distinctionbetween missense and nonsense mutations,however, is not always biologically relevant. Forexample, certain amino acid replacements maybe null alleles with regard to binding of thePAX3 protein to particular DNA sequences,such as sequences that regulate a modifier ofear function or development or both. In termsof deafness, these amino acid replacementswould be functionally equivalent to mutationsthat obliterate protein function.

All of the missense mutations are in one ofthe two DNA binding motifs, the paireddomain or the paired type homeodomain (fig1). A G test of likelihoods for deafness betweenamino acid replacements in the two domainsshows evidence for a significant difference(table 5A, Hi v H3, p=O.O14). Thus, amissense mutation in the paired domain ismore closely correlated with deafness than is amissense mutation in the paired type homeo-domain. This difference remains significanteven if possible effects of each unique allele areaccounted for (table 5A, H2 v H3, p=0.029).Nor can the heterogeneity be ascribed to theeffects of genetic background, since there is noevidence for sibship heterogeneity (table 5A,H2 v H4 and H3 v H4). Within each categoryof missense mutation (that is, either only thosefamilies segregating replacements in the paireddomain or only those families segregatingreplacements in the paired type homeodomain)there is also no evidence for sibship heteroge-neity (table 5B, C). There is, however, modestevidence for allelic heterogeneity among the

450

on Septem

ber 10, 2020 by guest. Protected by copyright.

http://jmg.bm

j.com/

J Med G

enet: first published as 10.1136/jmg.34.6.447 on 1 June 1997. D

ownloaded from

Deafness inWS4

StruCtIuc of PAX3

3 4107 151 196

32Pa ired dolinia

F45 LN47LP50 LR56LG81AG99D

451 586

5 6 7264 320 391

792Homeocdomiiain1

V265 FR27]i CR27 1 G

958 1173

Figure 1 Locations of amino acid replacements in PAX3. Numbers for the 3' most codons and the 3' most nucleotides ineach exon are listed at the exon boundaries above and below the ideogram, respectively. Shaded regions indicate the paireddomain and the paired type homeodomain. Missense mutations are listed below exons 2 and 6. All amino acid replacementsso far described occur in either the paired domain or the paired type homeodomain.

amino acid replacements in the paired typehomeodomain (table 5C, HI v H2, p=0.05).Thus, neither amino acid replacements in thepaired domain nor in the paired type homeodo-main show evidence for heterogeneity thatcould be the result of interactions with thegenetic background. However, the two groupsof missense mutations are significantly differ-ent from each other. When considering thisfinding, it is useful to bear in mind thatalternate transcripts of PAX3, which containthe paired domain but not the paired typehomeodomain, have been recovered from adulttissues.4' Whether similar transcripts are ex-pressed at critical stages of ear or braindevelopment is unknown. Also, the paireddomain and the paired type homeodomain actsynergistically to enhance binding to putativeconsensus sequences, and it is conceivable thatboth domains may be necessary for properbinding to certain sequences.4"" Finally, it ispossible that not all amino acid replacements inthe paired type homeodomain are equivalent,with some replacements showing a greaterinfluence of the genetic background on deaf-ness. The evidence for the latter is weak(p=0.05) given the multiple comparisonsperformed in these analyses, but cannot beentirely discounted. Until downstream targetgenes are identified we can only speculate onthe functional differences between paireddomain and paired type homeodomain mis-sense mutations.

Despite the apparent homogeneity of mis-sense and nonsense/frameshift mutations re-ported in table 4, tests of heterogeneity amongonly those pedigrees segregating nonsense/

Table 6 Tests for heterogeneity among pedigrees segregating protein truncating mutations(nonsense substitutions, frameshifts and in frame deletions)

Hypothesis InL No ofparameters

HI Penetrance of deafness same for all alleles -59.8 1H2 Penetrance of deafness different for each allele -46.4 14H3 Penetrance of deafness differs by exon location of the

mutation -56.9 6H4 Penetrance of deafness different in each sibship -23.8 45

Comparison G2 p (dl)HlvH2 26.8 0.013(13)Hl vH3 5.8 0.326 (5)HlvH4 71.9 0.005 (44)H2 v H3 21.0 0.007 (8)H2vH4 45.1 0.049(31)H3 v H4 66.2 0.004 (39)

frameshift mutations (table 6) give differentresults from similar tests performed amongpedigrees segregating missense mutations(table 5). Unlike the missense mutations, thereis no significant evidence for heterogeneityamong the protein truncating mutations whenthey are classified by the exon location of themutations (table 6, Hl v H3, p=0.326). Thereis heterogeneity among the protein truncatingmutations, but this can be attributed to differ-ences in genetic background (table 6, HI v H4,p=0.005) even when possible allele effects areaccounted for (table 6, H2 v H4, p=0.049). Inthe aggregate, the comparisons reported intable 6 indicate that the protein truncatingmutations are largely equivalent, and probablyobliterate protein function regardless of wherethe mutation occurs in the gene. The effect thatthis has on the aetiology of deafness, however,appears to be significantly influenced byparameters that vary from sibship to sibship. Asargued earlier, we suggest that this is aninfluence of genetic background.Given the high population prevalence of

autosomal recessive deafness,45 the coincidenceof carrier status for a recessive deafness geneand heterozygosity for a mutant PAX3 alleleshould not be uncommon. If the recessivedeafness locus is transcriptionally regulated byPAX3 it would be plausible that haploinsuffi-ciency of PAX3 could cause deafness in a het-erozygote for a mutant recessive deafnessallele. This is admittedly speculative, but thereis evidence that oligogenic inheritance contrib-utes to WS1 phenotypes,46 and that digenicinheritance is involved in the manifestation ofthe WS type 2 phenotype.47 4 In severalfamilies, WS2 cosegregates with markers bothon chromosome 1 and on chromosome 3.49 Inaddition, Asher et al26 have recently shown thatthe phenotypic expression of a mutant mousePax-3 allele is modified, possibly by severalgenes, in the background introduced in aninterspecies cross.

ConclusionThe present analyses suggest that the aetiologyof deafness in Waardenburg syndrome involvesthe interaction ofPAX3 mutations with geneticfactors. The evidence we have presentedsuggests that nonsense mutations are function-ally equivalent and that the non-random varia-

2Exon 1Codorn 29

[it 85

8

479

1-37

451

on Septem

ber 10, 2020 by guest. Protected by copyright.

http://jmg.bm

j.com/

J Med G

enet: first published as 10.1136/jmg.34.6.447 on 1 June 1997. D

ownloaded from

Morell, Friednman, Asher, Robbins

tion arises from genetic background, whilemissense mutations have an allele specificeffect on the aetiology of deafness. It seemsprobable that attempts to detect loci modifyinghearing status would be most profitable in sin-gle WS 1 families in which deafness is concen-trated in sibships. It is also likely that familiessegregating nonsense rather than missensemutations would be more informative. Suchanalyses would benefit from a comprehensiveaudiological evaluation in all subjects, includ-ing the unaffected parents, who might havesubclinical hearing impairments indicative ofcarrier status for recessive deafness genes.While the identification of modifier genesinfluencing the penetrance of deafness in WS 1families may be difficult, it should be reassur-

ing to know that they exist and that thevariability of deafness in WS1 is not a whollystochastic event.

This work was supported in part by research grant number RO 1DC01160 from the National Institute on Deafness and otherCommunication Disorders, National Institutes of Health (toJHA and TBF) and by grant MCB-9305846 from the NationalScience Foundation (to LGR). JHA died during completion ofthis paper.

1 Waardenburg PJ. A new syndrome combining developmen-tal anomalies of the eyelids, eyebrows and nose root withpigmentary defects of the iris and head hair with congenitaldeafness. Am Hum Genet 1951;3:195-253.

2 Mendelian inheritance in man - online database (program).Baltimore: Johns Hopkins University Press, 1996.

3 Partington MW. Waardenburg's syndrome and heterochro-mia iridum in a deaf school population. Can Med Ass J1964;90:1008-17.

4 Hageman MJ, Delleman JW. Heterogeneity in Waardenburgsyndrome. Am] Hum Genet 1977;29:468-85.

5 Bard LA. Heterogeneity in Waardenburg's syndrome. ArchOphthalmol 1978;96:1193-8.

6 Arias S. Genetic heterogeneity in the Waardenburgsyndrome. Birth Defects 1971 ;VII(4):87-101.

7 Liu XZ, Newton VE, Read AP. Waardenburg syndrome typeII: phenotypic findings and diagnostic criteria. Ans 7 MedGenet 1994;53:1-8.

8 da-Silva EO. Waardenburg I syndrome: a clinical andgenetic study of two large Brazilian kindreds, and literaturereview. Am Med Genet 1991;40:65-71.

9 Foy C, Newton V, Wellesley D, Harris R, Read AP. Assign-ment of the locus for Waardenburg syndrome type I tohuman chromosome 2q37 and possible homology to thesplotch mouse. Am_Hum Genet 1990;46:1017-23.

10 Morell R, Friedman TB, Moeljopawiro S, Hartono,Soewito, Asher JH Jr. A frameshift mutation in the paireddomain of HuP2 is responsible for Waardenburg syndrometype 1 in an Indonesian family. Hunm Mol Genet 1992;1:243-7.

11 Farrer LA, Arnos KS, Asher JH Jr, et al. Locus heterogene-ity for Waardenburg syndrome is predictive of clinical sub-types. Am_Hum Genet 1994;55:728-37.

12 Baldwin CT, Hoth CF, Amos JA, da Silva EO, Milunsky A.An exonic mutation in the HuP2 paired domain genecauses Waardenburg's syndrome. Nature 1992;355:637-8.

13 Baldwin CT, Hoth CF, Macina RA, Milunsky A. Mutationsin PAX3 that cause Waardenburg syndrome type I: ten newmutations and review of the literature. Am _7 Med Genet1995;58:115-22.

14 Hoth CF, Milunsky A, Lipsky N, Sheffer R, Clarren SK,Baldwin CT. Mutations in the paired domain of the humanPAX3 gene cause Klein-Waardenburg syndrome (WSIII)as well as Waardenburg syndrome type I (WSI). Anm ]fHunGeuiet 1993;52:455-62.

15 Lalwani AK, Brister JR, Fex J, et al. Further elucidation ofthe genomic structure of PAX3, and identification of two

different point mutations within the PAX3 homeobox thatcause Waardenburg syndrome type 1 in two families. Anm]Hum Genet 1995;56:75-83.

16 Morell R, Friedman TB, Asher JH Jr. A plus-one frameshiftmutation in PAX3 alters the entire deduced amino acidsequence of the paired box in a Waardenburg syndrometype 1 (WS I) family. Hum Mol Genet 1993;2:1487-8.

17 Morell R. Mutations responsible for Waardenburg syndronlestype 1 and 2. PhD dissertation, Michigan State University,1993.

18 Morell R, Carey ML, Lalwani AK, Friedman TB, Asher JHJr. Three mutations in the paired homeo domain of PAX3that cause Waardenburg syndrome type 1. Hum Hered (inpress).

19 Tassabehji M, Read AP, Newton VE, et al. Waardenburg'ssyndrome patients have mutations in the human homo-logue of the Pax-3 paired box gene. Nature 1992;355:635-6.

20 Tassabehji M, Read AP, Newton VE, et al. Mutations in thePAX3 gene causing Waardenburg syndrome type 1 andtype 2. Nat Genet 1993;3:26-30.

21 Goulding MD, Chalepakis G, Deutsch U, Erselius JR,Gruss P. Pax-3, a novel murine DNA binding proteinexpressed during early neurogenesis. EMBO _7 1991;1O:1135-47.

22 Stuart ET, Kioussi C, Gruss P. Mammalian PAX genes. In:Campbell A, Anderson W, Jones E, eds. Anniual review ofgenetics. Palo Alto: Annual Reviews Inc, 1994:219-36.

23 Pierpont JW, St Jacques D, Seaver LH, Erickson RP. A fam-ily with unusual Waardenburg syndrome type I (WSI), cleftlip (palate), and Hirschsprung disease is not linked toPAX3. Clin Genet 1995;47:139-43.

24 Moase CE, Trasler DG. Spinal ganglia reduction in thesplotch-delayed mouse neural tube defect mutant. Teratol-ogy 1989;40:67-75.

25 Moase CE, Trasler DG. Delayed neural crest cell emigrationfrom Sp and Spd mouse neural tube explants. Teratology1990;42: 171-82.

26 Asher JH Jr, Harrison RW, Morell R, Carey ML, FriedmanTB. Effects of Pax3 modifier genes on craniofacialmorphology, pigmentation, and viability: a murine modelof Waardenburg syndrome variation. Genonics 1996;34:285-98.

27 Woolf B. The log-likelihood ratio test (the G-test). Ann HumnGenet 1957;21:397-409.

28 Bishop YMM, Fienberg SE, Holland PW. Discrete multivari-ate analysis: theory and practice. Cambridge, MA: MITPress, 1975.

29 Edwards AWF. Likelihood. Baltimore: Johns HopkinsUniversity Press, 1992.

30 Terwilliger JD, Ott J. Handbook of human genetic linkage.Baltimore: Johns Hopkins University Press, 1994.

31 Asher JH Jr, Morell R, Friedman TB. Waardenburgsyndrome (WS): the analysis of a single family with a WS 1mutation showing linkage to RFLP markers on humanchromosome 2q. Am_

Hum Genet 1991;48:43-52.32 Tassabehji M, Newton VE, Read AP. Waardenburg

syndrome type 2 caused by mutations in the human micro-phthalmia (MITF) gene. Nat Genet 1994;8:251-5.

33 Asher JH Jr, Sommer A, Morell R, Friedman TB. Missensemutation in the paired domain of PAX3 causes

craniofacial-deafness-hand syndrome. Hunm Mutat 1996;7:30-5.

34 Sommer A, Young-Wee T, Frye T. Previously undescribedsyndrome of craniofacial, hand anomalies and sen-

sorineural deafness. Anm 7Med Genet 1983;15:71-7.35 Carezani-Gavin M, Clarren SK, Steege T. Waardenburg

syndrome associated with meningomyelocele. Ans ] MedGenet 1992;42: 135-6.

36 Baldwin CT, Lipsky NR, Hoth CF, Cohen T, Mamuya W,Milunsky A. Mutations in PAX3 associated with Waarden-burg syndrome type I. Hum Mutat 1994;3:205-11.

37 Tassabehji M, Newton VE, Liu XZ, et al. The mutationalspectrum in Waardenburg syndrome. Hunm Mol Genet1995;4:2131-7.

38 Hol FA, Hamel BCJ, Geurds MPA, et al. A frameshift muta-tion in the gene for PAX3 in a girl with spina bifida andmild signs of Waardenburg syndrome. ] Med Genet1995;32:52-6.

39 Kapur S, Karam S. Germ-line mosaicism in Waardenburgsyndrome. Clin Genet 199 1;39:194-8.

40 Farrer LA, Grundfast KM, Amos J, et al. Waardenburg syn-drome (WS) type 1 is caused by defects at multiple loci,one of which is near ALPP on chromosome 2: first reportof the WS consortium. Anm 7 Humn Genet 1992;50:902-13.

41 Tsukamoto K, Nakamura Y, Niikawa N. Isolation of twoisoforms of the PAX3 gene transcripts and their tissue-specific alternative expression in human adult tissues. HunGenet 1994;93:270-4.

42 Wilson D, Sheng G, Lecuit T, Dostatni N, Desplan C.Cooperative dimerization of paired class homeo domainson DNA. Genes Dev 1993;7:2120-34.

43 Underhill DA, Vogan KJ, Gros P. Analysis of the mouse

Splotch-delayed mutation indicates that the Pax-3 paireddomain can influence homeodomain DNA-binding activ-ity. Proc NatlAcad Sci USA 1995;92:3692-6.

44 Chalepakis G, Jones FS, Edelman G, Gruss P. Pax-3contains domains for transcription activation and tran-

scription inhibition. Proc Natl Acad Sci USA 1 994;91:12745-9.

45 Cohen MMJ, Gorlin RJ. Epidemiology, etiology, and geneticpatterns. In: Gorlin RJ, Toriello HV, Cohen MMJ, ed.Hereditary hearing loss and its syndronmes. Oxford: OxfordUniversity Press, 1995:9-21.

46 Pandya A, Xia XJ, Landa BL, et al. Phenotypic variation inWaardenburg syndrome: mutational heterogeneity, modi-fier genes or polygenic background? Hunm Mol Genet 1996;5:497-502.

47 Morell R, Pierpont J, Guo W, et al. A frameshift mutation inMITF co-segregates only with Waardenburg syndrometype 2 (WS2) in a family with both WS2 and ocularalbinism (OA) (abstract 1276). Am ] Huns Genet 1995;54:A22 1.

48 Morell R, Spritz RA, Ho L, et al. Apparent digenic inherit-ance of Waardenburg syndrome type 2 (WS2) andautosomal recessive ocular albinism (AROA). Hum MolGenet (submitted).

49 Lalwani AK, Baldwin CT, Morell R, et al. A locus forWaardenburg syndrome type II maps to chromosomeIp13.3-2.1.Anm7Hunt Genet 1994;55:A14.

452

on Septem

ber 10, 2020 by guest. Protected by copyright.

http://jmg.bm

j.com/

J Med G

enet: first published as 10.1136/jmg.34.6.447 on 1 June 1997. D

ownloaded from