use of hla marker associations and hla haplotype linkage to estimate disease risks in families with...

Clinical Genetics 1985: 28: 185-1 98

Use of HLA marker associations and HLA haplotype linkage to estimate disease risks in families with gluten-sensitive

enteropathy HENRY J. LIN, JEROME I. ROTTER AND WILLIAM J. CONTE

Division of Medical Genetics, Departments of Medicine and Pediatrics, UCLA School of Medicine, Harbor-UCLA Medical Center, Torrance, CA, U.S.A.

Based on a two-locus, double recessive model, we derive formulas for the risks that relatives of individuals with gluten-sensitive enteropathy (GSE) will also develop the disease. The calculations take advantage of: (i) the linkage between the HLA locus and one of the two proposed GSE loci, and (ii) the preferential association of the HLA-DR3 and DR7 alleles with the GSE disease allele that occupies the HLA-linked locus. We use Bayes’ rule to quantitate the strength of the association between the GSE disease allele and the HLA marker allele. This method predicts that siblings of the proband have an overall 10% risk for GSE, which is consistent with observed family data. This predicted risk rises to 30% when siblings are HLA-identical to the proband (also consistent with observed data) or when the sibling has the DR3 allele in the HLA haplotypes not shared with the proband. In those populations where DR7 also is associated with GSE, siblings of probands have a 10% predicted risk for GSE when only one HLA haplotype is shared with the proband and DR7 is included in the unshared haplotype. Other DR alleles are associated with much lower disease risks. By separating individuals into high and low risk groups, HLA typing identifies those individuals who would benefit from further diagnostic procedures. This general strategy should be applicable to other multilocus, marker-associated diseases.

Received 2 July 1984, revised, accepted for publication 18 June 1985

Key words: Association; disease risks; gene marker; gluten-sensitive enteropathy; HLA; linkage.

Gluten-sensitive enteropathy (GSE), also known as coeliac disease or coeliac sprue, is a familial disorder of malabsorption in adults and children. The pathologic change that leads to the malabsorption is atrophy of the small bowel villi resulting from the toxic effects of gluten, a constituent of wheat (Falchuk 1979). Though this environmental factor is well known, genetic susceptibility also appears to be essential, as demonstrated by the marked familial ag- gregation of the disorder and its high con-

cordance rate in monozygotic twins (Mylot- te et al. 1974, Ellis 1980, Stokes et al. 1973, Polanco et al. 1981).

Available evidence is most consistent with the hypothesis that the genetic predisposition to GSE is due to disease alleles at two unlinked loci (van Rood et al. 1975, Peiia et al. 1978, Strober 1980, Greenberg & Rotter 1981). The first locus was identified in the early 1970s when GSE was found to be associated with the histocompatibility antigen HLA-B8 (Falchuk et al. 1972, Sto-

186 L I N E T A L .

kes et al. 1972). This antigen was detected in some 80% of GSE patients, compared to its 20% prevalence in controls. Family studies showed that GSE usually segregated with the B8 marker allele (Albert et al. 1973, Harms et al. 1974, Robinson et al. 1980). When HLA-D locus typing became available, the DR3 antigen proved to have a stronger association with the disease than B8 (Keuning et al. 1976), and the HLA- DR7 antigen was identified as an important GSE-associated antigen among southern Europeans (Polanco et al. 1981). To illus- trate, the prevalences of the HLA-DR3 and DR7 markers among patients with GSE and among normal individuals in Spain and in The Netherlands are listed in Table 1 (Peiia et al. 1981, Mearin et al. 1983).

Though it was crucial in establishing a genetic basis for the disease, discovery of the HLA-linked locus could not explain the observed patterns of inheritance of GSE. Only a fraction of HLA-identical siblings of GSE patients have GSE (Peiia et al. 1981). In addition, families have been reported in which GSE segregated independently of HLA markers (Falchuk et al. 1978). New insight for the mode of inheritance of GSE came when B-cell alloantigens specific for GSE were found using the sera of mothers of GSE patients (Mann et al. 1976). Shortly thereafter, these B-cell alloantigens were shown to segregate independently of the HLA complex, and a second GSE locus encoding these specific B-cell alloantigens, together with an HLA-linked susceptibility

factor, were proposed to provide susceptibility to the disease (Peiia et al. 1978).

The original two-locus model proposed that the DR3 marker allele was always and only found with the HLA alleles associated with GSE, i.e. that either DR3 itself was contributory or it was in near absolute linkage disequilibrium with a disease allele at a nearby GSE locus (Peiia et al. 1978). The HLA-linked disease susceptibility gene was proposed to be dominant, because only a single dose of the DR3 allele appeared to be present in some affected individuals. The GSE-associated B-cell alloantigen allele appeared to be recessively expressed, because the B-cell alloantigens were uniformly present in both healthy parents of a series of GSE children. The model could thus be cal- led dominant-recessive. For the HLA-linked locus, the disease allele frequency was estimated to be 0.19, and for the non-HLA locus, it was estimated to be 0.09 (Peiia et al. 1978).

Several lines of reasoning argue that recessive expression of disease alleles at both loci is more likely (Greenberg & Rotter 1981, Greenberg et al. 1982). First, the disease prevalence is much lower than predicted by the dominant-recessive model and its proposed allele frequencies. In contrast, the calculated disease prevalence based on these allele frequencies in a double recessive model is quite consistent with the observed disease prevalence (Greenberg & Rotter 198 1). Secondly, the expected proportion of siblings of probands who are affected ac-

Table 1

Prevalence of HLA-DR3 and DR7 markers in GSE patients and in controls

GSE Patients (FMD) Controls (FMP) DR3 DR7 DR3 DR7

The Netherlands (Peiia et al. 1981) 0.96 0.21 0.22 0.23 SDain (Mearin et al. 1983) 0.71 0.61 0.17 0.38

FMD stands for frequency of the marker in the diseased. FMP stands for frequency of the marker in the population.

H L A , L I N K A G E , A S S O C . & D I S E A S E R I S K I N G S E 187

cording to a double recessive model agrees with the percentage observed, whereas an estimate based on a dominant-recessive model is much too high (Greenberg & Rotter 1981). Formal segregation testing by likelihood methods confirmed this result (Greenberg & Lange 1982). The dominant- recessive model can be made compatible with the population prevalence by adjusting the disease allele frequencies, but this does not improve the model's prediction of the population segregation ratio (Greenberg & Rotter 1981). Thirdly, HLA marker frequencies were consistent with recessive inheritance of susceptibility at the HLA locus and incompatible with dominant inheritance (Greenberg et al. 1982). Recessive inheritance at this locus, however, may still be an oversimplification (Mearin et al. 1983, DeMarchi et al. 1983).

Based on its ability to accurately predict both the population prevalence and segregation ratio, the double recessive model is currently the best overall approximation of the mode of inheritance of GSE. Herein, we utilize this model to derive equations for estimating risks for disease to relatives of GSE patients. These equations further characterize the double recessive model and present new ways to corroborate it. In addition, these formulations provide an important new tool for genetic counseling and screening of relatives at risk by indicating those who should undergo intensive screening, such as small bowel biopsy and gluten challenge. This proposed strategy for the simultaneous use of HLA linkage and association data is, moreover, a framework for identifying individuals at high risk not only for GSE, but for other multilocus diseases as well.

Material and Methods

Terms and Notation The method and results will use the follow-

ing notation. Dyshomozygosity, homozygosity for a disease allele at a single locus; GSEI, the HLA-linked and HLA-associated GSE disease locus on chromosome six; GSE2, the proposed B-cell alloantigen-associated GSE disease locus; D1, the disease allele for the GSEl locus; 0 2 the disease allele for the GSE2 locus; M , a specific HLA marker allele linked to the GSEl locus (M' is the HLA marker allele on the individual's homologous chromosome six); p(D1) = the frequency of D1; p(D2) = the frequency of D2; P(D 1 ,D 1) = the probability for homozygosity for the disease allele at GSE1; P(D2,D2)=the probability for homozygosity for the disease allele at GSE2; p(M)=the frequency of the marker allele denoted by M; p(MlD1) = the probability that a chromosome carries the marker allele M when it is known that it carries D1; p(D1 IM)= the probability that a chromosome carries D1 when it is known that it carries the HLA marker allele M.

General Formulation Our object is to determine the probability that a relative of a patient with GSE is homozygous 'for disease alleles at the two unlinked loci, which we call GSEl and GSE2, following earlier nomenclature (Po- lanco et al. 1981). GSEl is the HLA-linked locus and GSE2 the B-cell alloantigen locus.

The clinical problem and major assump- tions of our model are portrayed in pedigree form in Figure 1. The proband (11-7) is homozygous for disease alleles both at the HLA-linked and at the B-cell alloantigen loci, and therefore has GSE. His parents (I- 1 and 1-2) and offspring (111-5) are obligate heterozygotes at both loci, but might be homozygous at either locus, depending on the disease allele frequencies and inheritance of these disease alleles from individuals marrying into the family, i.e. the population at large. The siblings (11-1, 11-3, 11-5) can be normal, heterozygous at one or

188 L I N E T A L .

I ’*

locus locus GSE 1 HETEROZYGOTE

locus locus GSE 1 HETEROZYGOTE GSE 2 HETEROZYGOTE

IG*l [ Fig. 1 . Approach to determining the risk for GSE to relatives of an affected proband assuming two-locus, double recessive inheritance. The proband (arrow) is homozygous for disease alleles at both GSE loci. His parents and offspring are obligate heterozygotes at both GSE loci. Some of the family’s disease alleles at GSEl are easily identified by testing for HLP. haplotypes known to be linked to a disease allele in the proband. All other disease alleles can be assigned only probabilistically, as discussed in the text.

both loci, or homozygous for a disease allele at one or both loci. Obligate heterozygotes (11-3) and homozygotes (11-5) for the disease allele at GSEl in the sibship are easily identified by HLA typing. Any of the nieces and nephews (111-2 through 111-4), who are direct descendants of obligate carriers, are also at increased risk for the disease.

Probabilities that these relatives are homozygous for a disease allele at the GSEl or GSE2 locus are derived below. We first assume no marker information, such as HLA types or B-cell alloantigen types, is available, and the derivations are based only on Mendelian segregation and the Hardy- Weinberg law. We then use the information that one disease locus is linked to the HLA complex and has a disease allele that is strongly associated with specific HLA marker alleles, namely DR3 and DR7. This information is applied by a method reported earlier for screening families with recessive, marker-linked, and marker-associated diseases, such as hereditary hemochromatosis

(Conte & Rotter 1984), but simplified by the use of Bayes’ rule (Lin et al. 1984). Siblings who are HLA-identical to the proband are presumed homozygous for a disease allele at the HLA-linked locus. Rela- tives who share one HLA haplotype with the proband are obligate heterozygote carriers of a disease allele at the HLA-linked locus. The probability that such relatives have a second HLA-linked disease allele de- pends on the strength of the association between the individual’s unshared HLA marker and the disease allele. The probability is quantitated by Bayes’ rule (Lin et al. 1984). GSE B-cell alloantigen testing is not widely available and cannot be routinely applied to clinical problems. Because GSEl and GSE2 are independent loci, the risk for homozygosity for a disease allele at both loci simultaneously is obtained by multiplying the probabilities for homozygosity for a disease allele at the individual loci.

As a final prefatory note, we introduce the term, dyshomozygous, as an alternative to the phrase, homozygous,for a disease allele. The analogous term of simple Mendelian genetics, uflected, will not always apply in multilocus disorders. In other words, the individual who is “affected” at a given locus will not always be affected with the disease (or the disease susceptibility). The term, dyshomozygous, succinctly captures the idea of an individual having a pair of disease alleles. This and other terms may be needed as we advance in our knowledge of multilocus diseases (Cotterman 1983, Rotter & Landaw 1984).

Dyshomozygosity at the GSEl locus and at the GSE2 locus when markers are unknown: The probabilities for dyshomozygosity at the GSEl locus, when HLA types are unknown, and at the GSE2 locus are derived from Mendel’s rules of segregation and the Hardy-Weinberg law. For now, let


P = probability that the relative is homozygous for the GSE allele at either locus

p = disease allele frequency in the general population for either locus.

When we need to be more specific, we will use the notation P(D1,Dl) and P(D2,D2) for the probabilities for dyshomozygosity at the GSEl and GSE2 loci, respectively, and the notation p(D1) and p(D2) for the disease allele frequencies for the GSEl and GSE2 loci, respectively. In our calculations, p(D 1) = 0.19, and p(D2) = 0.09, as originally proposed (Peiia et al. 1978) and supported by other analyses (Greenberg & Rotter 1981, Greenberg et al. 1982).

( I ) Parents and oflspring: These relatives are obligate heterozygotes and comprise the simplest case. The probabilities that they are dyshomozygous (i.e. the probabilities of having a second disease allele) are equal to the frequencies of the disease allele. In other words, for parents and offspring

P(Dl,Dl)=p(Dl) and P(D2,D2) =

(Equation 1)

( 2 ) Siblings: The probability for dyshomozygosity for a sibling is the product of the probabilities that each parent transmits a disease allele.

P = probability that the father transmits a disease allele x probability that the mother transmits a disease allele

= (1/2+ 1/2p) x (1/2+ 1/2p).

The 1/2p terms account for the possibility that the parents may be homozygotes for a disease allele at one or the other locus. This is possible even in unaffected individuals, because heterozygosity at one locus masks such homozygosity at the other locus when both disease alleles are recessively ex-

pressed. If the disease allele frequencies are low, the probability that a parent transmits a disease allele simplifies to 1/2. Factoring and multiplying,

P = 1/2(1 +p) x 1/2(1 +P> = 1/4(1+p)’.

(If one of the parents has GSE, then that parent’s chance of passing on the disease allele is unity rather than 1/2(1 +p).) So, for siblings of the proband

P(D13D1)= 1/4[1 +P(D1)12 (Equation 2) P(D2,D2) = 1/4[1 +p(D2)I2.

( 3 ) Nieces and nephews: The probabilities for dyshomozygosity for these family members are the products of the probabilities that each of their parents transmits a disease allele. This probability is equal to the disease allele frequency, p, for the parent who marries into the family. The other parent is a sibling of the proband and has a different probability of transmitting a disease allele. Let

p’ = probability that the proband’s sibling transmits a disease allele

= probability that the proband’s sibling is homozygous for a disease allele+ 1/2 (probability that the proband’s sibling is heterozygous)

= 1/4(1 + P ) ~ + 1/2{2[1/2(1 +p)] x [ I - 1/2(1 +P)l).

Note that the first term was derived above. The term within the large brackets, { }, can be thought of as 2d(l-d), where d is the probability that a given parent of the proband transmits a disease allele. Then ( 1 4 ) is the probability that the other parent does not. The factor 2 enters because of the reci- procity of the mating types. Continuing,

p’ = 1/4(1 +p>’+ 1/2(1 +p) x 1/2(1 -p)

190 L I N E T A L .

= 1/4(1 +p)'+ 1/4(1 +p) (1 -p) = 1/4(1+p) ( l + p + l - p ) = 1 / 4 ( 1 + p ) ~ 2 = 1/2(1+p).

Then

P = p x p ' = 1/2p(l+p).

Thus, for nieces and nephews,

P(DI,Dl)= 1/2p(Dl) [l+p(Dl)]

P(D2,D2)= 1/2p(D2) [l +p(D2)]. (Equation 3)

Dyshomozygosity at the GSEl locus when HLA markers are known: The probability for homozygosity for a disease allele at the GSEl locus can be refined by HLA typing. Finding an HLA haplotype identical to one of those of the proband immediately de- monstrates the presence of the GSEl disease allele in a relative. In addition, a DR3 or DR7 marker allele, even though not part of either of the proband's haplotypes, has a higher probability of traveling with an otherwise unsuspected GSE gene, as is shown below. Other markers have a lower probability.

Bayes' rule is used to quantitate the probability that a disease gene is associated with

p(M) = marker allele frequency in the general population

= 1 - (1 - FMP)i/2, where FMP is the frequency of the marker in the population.

p(M/Dl) = probability that a chromosome carries the marker when it is known that it carries the GSEl disease allele

= 1 - (1 - FMD)I/*, where FMD is the frequency of the marker in the diseased.

Values for FMP (controls) and FMD (GSE patients) are in Table 1. The derivation of p(M) and p(MlD1) has been previously de- scribed (Lin et al. 1984). The frequency of the HLA-linked disease allele, p(Dl), is 0.19, as discussed above.

The probability that a relative is homozygous for a GSEl disease allele is obtained by applying Bayes' rule to both of that relative's HLA markers:

P(D1,D 1) = p(D1 IM)" p(D1 IM')", (Equation 5)

where M is an HLA marker encoded on one chromosome, M' is the corresponding HLA marker allele on the other chromosome and

specific markers (Feller 1968, Hoe1 et al. 1971). The probability that a GSEl disease allele is present when a designated marker allele is present (p(D1 IM) is calculated from Bayes' rule according to the formula

0, if the marker (M or M') is identical to one of the proband's markers 1, if the marker (M or M') is not n =

identical to one of the proband's markers.

(Equation 4) Overall disease risk: We now have all the equations needed to estimate the risk for dyshomozygosity at the GSEl and GSE2

p(DI) = GSEl disease allele fre- loci to parents, offspring, siblings, and nieces or nephews of patients with GSE. These equations are summarized in Table 2. The

where

quency in the general POPU-

lation,


Table 2

Formulas for probabilities for homozygosity for disease alleles at each of the two GSE loci for relatives of GSE patients

Locus Parent or offspring Sibling Niece or nephew

GSEI 2 shared HLA haplotypes” 1 - - 1 shared HLA haplotype p(D1 IM) p(D1IM) P(DI IM) No shared HLA haplotype - p(D1 IM)p(DI IM’) p(D1 IM)P(DIIM’) Unknown HLA P(W 1/4 [ I +p(D1)I2 112 P(D1” +P(Dl) l

GSE2 p(D2) 114 [ I + p(D2)I2 112 P ( W [ I +p(D2)1

a This remark indicates the number of HLA haplotypes the relative shares with the GSE proband. Notation is explained in Material and Methods under “Terms and Notation.”

overall risk for GSE P(D1,Dl) x P(D2,D2).

is the product,

Table 3

Probabilities for homozygosity for a disease allele for relatives of GSE patients at GSEI, GSES, and both loci simultaneously, when HLA

haplotypes are unknown

Parent or Niece or Locus offspring Sibling nephew

GSEI 0.19 0.35 0.1 1 GSE2 0.09 0.30 0.049 Both locis 0.017 0.10 0.005

a Probability of homozygosity at both loci is the product of the first two terms in each column.

Results

Our calculations of the risks for GSE to relatives of affected individuals are presented in stepwise fashion in Tables 3 through 6. Table 3 shows the risks for homozygosity for a disease allele, or dyshomozygosity, at each of the two GSE loci, without considering HLA types. Parents and offspring of individuals with GSE have a 10% to 20% chance of dyshomozygosity at either locus. Siblings have roughly a 30% risk at each locus, and nieces and nephews have a 5% to 10% risk. By multiplying the risk for dyshomozygosity at one locus by the risk for dyshomozygosity at the other

Table 4

Risks that designated HLA marker alleles have an attendant GSEl disease allele (p(D1JM)) as calculated by Bayes’ rulea

The Netherlands Spain Other Other

HLA marker Risk Bayes’ rule terms Risk Bayes’ rule terms allele, M P(D1lM) P(M) P(MlD1) P(D1 IM) P(M) P(M I D1)

~~

DR3 I b 0.12 0 80 DR7 0.17 . 0.12 0.1 1 DRn0n3,7~ 0.022 0.76 0.09 DR(any)d 0.19 1 1

0.98 0.089 0.46 0.34 0.21 0.38 0.043 0.70 0.16 0.19 1 1

See Equation 4; p(D1)=0.19. The notation is explained in Material and Methods under “Terms and Notation.” The calculated figure is actually greater than 1. DRnon3,7=any DR allele except DR3 and DR7. DR(any)=any DR allele.

192 L I N E T A L

Table 5

Probabil i t ies for homozygosi ty f o r a d isease a l le le a t the GSEl locus for re la t ives of GSE patients wi th the des ignated HLA marker haplo-

types

HLA markers The Nether- M M' lands

shared shared 1 DR3 shared 1 DR7 shared 0.17

DRnon3,7 shared 0.022 any unshared shared 0.19

DR3 DR3 1 DR7 DR3 0.17

DRnon3,7 DR3 0.022

DR7 DR7 0.029 DRnon3,7 DR7 0.004

any unshared DR7 0.032

DRnon3,7 DRnon3,7 <0.001 any unshared DRnon3,7 0.004

any unshared any unshared 0.036

any unshared DR3 0.19

Spain

1 0.98 0.34 0.043 0.19

0.96 0.33 0.042 0.19

0.12 0.015 0.065

0.002 0.008

0.036

Shared =an HLA marker haplotype identical to one of the proband's HLA haplotypes. Any unshared = an HLA marker haplotype not identical to one of the proband's. Other notation is as in Table 4.

locus, we obtain an overall risk for disease. The overall risks listed in Table 3 could be used for counseling families with GSE before HLA typing is done. In the case of siblings of affected probands, for example, this model predicts a 10% risk for GSE, which is in close accord with published studies (Ellis 1980).

When the HLA association data are in- corporated into the two-locus, double recessive GSE model, the predicted risks for disease are modified. We show this for populations in two different countries, Spain and The Netherlands, which were chosen because there is primarily only an HLA-DR3 association in one, whereas in the other both DR3 and DR7 marker alleles are prominently associated with GSE (Table 1). Using Equation 4, the DR3 marker allele has nearly a 100% risk of being linked to a GSE disease allele in both populations (Ta-

ble 4). Similarly, our calculations indicate that the DR7 marker allele carries a 34% risk in Spain and a 17% risk in The Nether- lands of being linked to a GSE disease allele. In these two populations, the marker-related adjustments separate the probabilities for dyshomozygosity at the HLA-linked locus into high, medium, and low ranges of about loo%, 10% to 30%, and a few per- cent or less, respectively (Table 5).

The best overall disease risk estimates that could be given to families seeking counseling for GSE use these HLA-related adjustments and are listed in Table 6. The highest risks for GSE are predicted to be 9% in certain parents and offspring, 30% in specific siblings, and 5% in the high risk nieces and nephews of affected individuals. Such high risks are present when the individual in question shares both HLA haplotypes with the proband, or has the DR3 marker allele in the unshared HLA haplotypes. With any other pairings of HLA marker haplotypes, including haplotypes con- taining a DR7 allele, the risks for GSE are 5% to 10% or less.

The incidences of GSE in relatives of affected individuals observed in published studies are compared to the predicted risks for these relatives in Table 7. For most groups of relatives, the predicted risks are very close to the observed risks. The 1.7% predicted risk for parents of affected individuals is close to the incidence of less than 4% reported in three studies. The overall 10% predicted risk for siblings of an affected proband regardless of HLA type is at the midpoint of the 6% to 13% reported by most studies. The 30% predicted risk for siblings who share both HLA haplotypes with the affected proband approximates the incidence of GSE in such siblings reported in two studies. However, the predicted risk for disease to offspring of individuals with GSE is 1.7%, but the reported incidence is in the range of 6% to 10%. In this latter


Table 6

Overal l r isks for GSE for relat ives of GSE patients w i th the designated HLA marker haplotypes

Niece or M M' offspring Sibling nephew offspring Sibling nephew

HLA markers The Netherlands Spain Parent or Niece or Parent or

shared DR3 DR7

DRnon3,7

unshared

DR3 DR7

DRnon3,7

unshared

DR7 DRnon3,7

any unshared

DRnon3,7 any

unshared

any unshared unknown

any

any

shared - shared 0.09 shared 0.015 shared 0.002

shared 0.017 DR3 - DR3 - DR3 -

DR3 - DR7 - DR7 -

DR7 -

DRnon3,7 -

DRnon3,7 -

unshared - unknown . 0.017

any

0.30 0.30 0.051 0.007

0.057

0.30 0.051 0.007

0.057

0.009 0.001

0.010

<0.001

0.001

0.01 1 0.10

- - 0.049 0.088 0.008 0.031 0.001 0.004

0.009 0.017

0.049 - 0.008 - 0.001 -

0.009 - 0.001 -

< 0.001 -

0.002 - <0.001 -

0.001 -

0.002 - 0.005 0.017

0.30 0.29 0.10 0.013

0.057

0.29 0.10 0.013

0.056

0.035 0.004

0.019

< 0.001

0.002

0.01 1 0.10

- 0.048 0.017 0.002

0.009

0.047 0.016 0.002

0.009

0.006 10.001

0.003

<0.001

< 0.001

0.002 0.005

See Tables 4 and 5 for explanations of marker designations.

situation, unrecognized ascertainment biases (such as an affected child increasing the probability of a family coming to attention) could have substantially raised these reported observed risks.

Discussion

This paper brings together four previously hypothesized concepts of the genetics of gluten-sensitive enteropathy. These concepts are: (i) two independent loci provide susceptibility to the disease, (ii) one locus is linked to the HLA region on chromosome six, (iii) the disease allele at the HLA-linked locus is preferentially associated with specific HLA alleles (DR3 and DR7), and (iv) the disease alleles at both loci are recessively expressed (Peiia et al. 1978, Strober 1980, Green- berg & Rotter 1981).

Using disease allele frequencies proposed by Peiia et al. (1978), we have elaborated a two-locus model that produces estimates of the risks that relatives of patients with GSE will also develop the disease if exposed to gluten. Our predictions of the incidence of GSE among parents and siblings of patients with GSE are identical to the observed incidences of these relatives. The approach is thus a potentially useful genetic counseling tool.

For offspring our predicted risks for GSE appear to be lower than the observed incidences. There are several possible reasons for this. First, family studies are difficult because of both practical problems and problems related to the nature of the disease: selection of families tends to be non- random, family members are often unavail- able for investigation, and accurate diag-

194 L I N E T A L .

Table 7

Observed risks for GSE for relatives of GSE patients as determined by small bowel biopsy studies

Reference

Number of Number of New relatives relatives GSE Observed

risk, %" available biopsied cases

Parent of proband HLA markers unknown Predicted risk=l.7%

MacDonald et al. 1965 34 12 0 0 Robinson et al. 1971 44 29 1 2.2-3.4 Mylotte et al. 1974 54 27 1 1.8-3.7 Shipman et al. 1975 64 53 4 6.2-7.5 Stokes et al. 1976 82 41 5 6.1-12 Rolles et al. 1981 45 45 3 6.7

Sibling of proband HLA markers unknown Predicted risk= 10%

MacDonald et al. 1965 63 31 4 6.3-1 3 Mylotte et al. 1974 142 80 9 6.3-1 1 Shipman et al. 1975 80 78 10 12-13 Stokes et al. 1976 143 81 18 13-22 Rolles et al. 1981 72 72 4 5.6

Offspring of proband HLA markers unknown Predicted risk= 1.7%

MacDonald et al. 1965 35 17 2 5.7-12 Stokes et al. 1976 101 60 12 12-20 Rolles et al. 1981 26 26 2 7.7

Sibling of proband HLA identical

Predicted risk=30% Peiia et al. 1981 9 5 3 33-60 Mearin et al. 1983 1 3b ? 4 31

Sibling of proband Half HLA identical; any DR allele in the unshared haplotype

Predicted risk=5.7% Peiia et al. 1981 13 4 3' 23-75 Mearin et al. 1983 294 ? 1 3.4

Sibling of proband No HLA haplotype shared with the proband

Predicted risk= 1.1 % Peiia et al. 1981 11 2 0 0 Mearin et al. 1983 17 ? 0 0

Sibling of proband Half HLA identical: DR3 allele in the unshared haplotype

Predicted risk=30% Mearin et al. 1983 3" ? 1 33

Sibling of proband Half HLA identical; DR7 allele in the unshared haplotype

Predicted risk=10% in Spain Mearin et al. 1983 3' ? 0 0

When a range is given for the observed risk, the lower limit is determined by dividing the number of new GSE cases by the number of relatives available, and the upper limit is determined by dividing the number of new GSE cases by the number of relatives biopsied.


nosis is difficult (Meeuwisse 1970). Biopsy of the small intestine is indispensable for the latter, because an individual's symptoms or the results of non-invasive tests do not reliably indicate the mucosal architecture. (A flattened mucosa, on the other hand, is not always diagnostic of GSE (Katz & Grand 1979)). Our analysis must reflect these difficulties. Several studies used for comparison to our predictions reported small bowel biopsies from most of the family members, but the selection of these families may not have been random. Fa- milies in which more than one member is affected, for example, are much more likely to be ascertained.

A second possible reason for the oc- casional differences between our predicted risks for GSE and the actual incidence of GSE among offspring of patients is inaccu- racy in the estimation of the disease allele frequencies. The frequency of 0.19 for the disease allele at the HLA-linked locus and the frequency of 0.09 for the disease allele at the proposed B-cell alloantigen locus are consistent with the GSE prevalence of one per 6,000 to one per 2,000 found in several parts of Great Britain (Greenberg & Rotter 198 1). The prevalence, however, varies widely from one geographic area to another, depending on both genetic and environmental factors (Mylotte et al. 1973, Stokes et al. 1976). In Western Ireland, for example, the disease prevalence may be as

high as one per 303 (Mylotte et al. 1973). This would necessitate both disease alleles having a substantially higher frequency, producing corresponding higher risk to relatives. This has indeed been observed (McKenna et al. 1983). With the appropri- ate population data, the special risks in the Irish population could also be enumerated by applications of the method presented in the paper.

If selection biases and variations in gene frequency do not account for all of the dis- crepancy between the observed incidences of GSE among relatives of affected individuals and our predictions, other reasons must be investigated. The two-locus model most likely requires refinement. There may be more than one disease allele for the HLA-linked locus (which we call GSEl). Two observations in favor of this are: (i) the frequency of the DR3/DR7 genotype is increased over the other HLA genotypes in affected persons (Mearin et al. 1983, De- Marchi et al. 1983, Tiwari et al. 1984), and (ii) the DR7 marker was present in 57% of individuals who developed the disease before age 20, compared to its 26% prevalence among those who developed GSE la- ter in life, according to a recent study by Ellis et al. (1984). Heterogeneity of GSE genotypes, especially heterogeneity related to different ages of onset, has not been in- corporated into any genetic model.

Until more accurate genetic models of

Families with monozygotic twins were excluded from this tabulation (Families 5, 25, 30, and 68 of Mearin et al. 1983). Also excluded were families in which homozygosity for DR marker alleles in one parent or congruence of DR marker phenotypes in the mother and father made assignment of shared and unshared haplotypes among the siblings ambiguous (Families 3, 6, 12, 17, 19, 27, 28, 29, 31, 32, 33, 44, 45, 51, 52, 55, 60, 63, and 66 of Mearin et al. 1983). One of these affected siblings is in a family in which a parent is affected (Family 16 of Pefia et al. 1981). This sibling shares one marker haplotype with his affected sister and another marker haplotype with his affected father. Families 3, 8, 9, 27, 32, 43, 65, and 67 of Mearin et al. 1983, were excluded because of the problem with ambiguity mentioned in Footnote b. Families 3, 8, 9, 12, 19, 27, 31, 45, 52, 55, 63, and 68 of Mearin et al. 1983, were excluded because of the problem with ambiguity mentioned in Footnote b.

' Families 6,32,63, and 68 of Mearin et al. 1983, were excluded because of the problem with ambiguity mentioned in Footnote b.

196 L I N E T A L .

GSE can be constructed or reliable screening tests become available (Burgin-Wolff et al. 1983, O’Farrelly et al. 1983), the investigation of families with the disease will re- main a difficult matter. Others have recently commented on the value of HLA typing the siblings of probands to look for haplotypes known t o be linked to the disease gene (McKenna et al. 1983). The model and method proposed herein extend this approach to classes of relatives other than HLA identical siblings. This model, which makes use of both HLA linkage and specific HLA associations in GSE, can be used to quantitate risks to asymptomatic family members. Because these risks, with the data available, fall naturally into high (30%) and low ranges (5% or less), HLA typing may be helpful in identifying those individuals more likely to benefit from invasive and time-consuming diagnostic tests, such as small bowel biopsy with and without gluten challenge, and those individuals in whom the risks of disease are low enough that it would be safer to observe them clinically. The early diagnosis of GSE would thereby be made more systematic and cost effective. As new screening and diagnostic tests that could be applied periodically to relatives at risk become available, the single application of these genetic marker studies and the resulting risk assessment could then help identify those relatives to whom these diagnostic methods should be applied, and with what frequency.

Finally we would like to suggest that a fundamental aspect of the model proposed for GSE may become a recurring theme in the genetics of other diseases, namely multilocus inheritance. Transmission of disease via a small number of genetic loci is being recognized with increasing frequency as new marker systems are explored (see Rotter & Landaw 1984, for a brief list of such diseases). The multilocus approach has the potential to provide the basis for eluci-

dating the etiologies of disease that are now poorly understood. This is especially true for diseases that show a familial or ethnic clustering but do not follow classic Mendeli- an inheritance. While the list of such disorders for which multiple disease loci have been clearly implicated is still short, the number is likely to increase dramatically as recombinant DNA techniques are applied. Thus, the concepts regarding identification of relatives at risk developed here for GSE are likely to have a much broader application for many diseases, both common and rare.

Acknowledgments

We would like to thank Dr. Gloria Petersen for her review of this manuscript, and Dau- ne Thorington for secretarial assistance.

Supported in part by NIH grants AM 33329, AM 17328, AM 36200 and A1 19340.

References

Albert, E. D., K. Harms, R. Wank, 1. Steinbauer- Rosenthal & S. Scholz (1973). Segregation analysis of HL-A antigens and haplotypes in 50 families of patients with coeliac disease. Transplant. Proc. 5, 1785-1789.

Burgin-Wolff, A,, R. M. Bertele, R. Berger, H. Gaze, H. K. Harms, M. Just, S. Khanna, K. Schurmann, E. Signer & D. Tomovic (1983). A reliable screening test for childhood coeliac disease: fluorescent immunosorbent test for gliadin antibodies. J . Pediutr. 102, 655-660.

Conte, W. J. & J. I. Rotter (1984). Use of association data to identify family rnernebers at high risk for marker-linked diseases. Am. J . Hum. Genet. 36, 152-1 66.

Cottennan, C. W. (1983). Relationship and probability in Mendelian populations. Am. J . Med. Genet. 16, 393440.

DeMarchi, M., A. Carbonara, N. Ansaldi, B. Santini, C. Barbera, I. Borelli, P. Rossino & S. Rendine (1983). HLA-DR3 and DR7 in coeliac disease: irnmunogenetic and clinical aspects. Gut 24, 70G712.

Ellis, A. ( 1 980). Coeliac disease: family and twin studies and disease associations. In Genetics


and Heterogeneity of Common Gastrointestinal Disorders J. I. Rotter, I. M. Samloff & D. L. Rimoin (eds.) New York, Academic Press, pp.

Ellis, A,, C. J . Taylor, M. Dillon-Remmy, J. C. Woodrow & R. B. McConnell (1984). HLA- DR typing in coeliac disease: evidence for genetic heterogeneity. Br. Med. J . 289, 1571-1573.

Falchuk, Z. M. (1979). Update on gluten-sensitive enteropathy. Am. J . Med. 67, 1085-1096.

Falchuk, Z. M., G. N. Rogentine & W. Strober (1 972). Predominance of histocompatibility antigen HL-A8 in patients with gluten-sensitive enteropathy. J . Clin. Invest. 51, 1602-1605.

Falchuk, Z. M., A. J. Katz, H. Schwachman, G. N. Rogentine & W. Strober (1978). Gluten- sensitive enteropathy: genetic analysis and or- gan culture study in 35 families. Scarid. J . Gas- troent. 13, 839-843.

Feller, W. (1968). An Introduction to Probability Theory and its Applications, Vol. 1, 3rd Edit. New York, John Wiley and Sons, Inc.

Greenberg, D. A. & J. I. Rotter (1981). Two locus models for gluten sensitive enteropathy: population genetic considerations. Am. J . Med. Genet. 8, 205-214.

Greenberg, D. A. & K. L. Lange (1982). A maximum likelihood test of the two locus model for coeliac disease. Am. J . Med. Genet.

Greenberg, D. A,, S. E. Hodge & J. I. Rotter (1982). Evidence for recessive and against dominant inheritance at the HLA-“linked” locus in coeliac disease. Am. J . Hum. Genet. 34,

Harms, K., G. Granditsch, E. Rossipal, H. Lud- wig, Z. Polymenidis, S. Scholz, R. Wank & E. D. Albert (1974). HL-A in patients with coeliac disease and their families. In Coeliac Disease (Proceedings of the 2nd International Coeliac Symposium) W. Th. J. M. Hekkens & A. S. Peiia (eds.) H. E. Stenfert Kroese B. V., Leiden,

Hoel, P. G., S. C. Port & C. J. Stone (1971). Introduction to Probability Theory. Boston, Houghton Mifflin Co.

Katz, A. J. & R. J. Grand (1979). All that flattens is not “sprue”. Gastroenterology 76, 375-317.

Keuning, J . J., A. S. Peiia, A. van Leeuwen, J. P. van Hoof & J. J. van Rood (1976). HLA-Dw3 associated with coeliac disease. Lancet i , 506-508.

Lin, H. J., W. J. Conte & J . I. Rotter (1984). Disease risk estimates from marker association data. Application to individuals at risk for hemochromatosis. Clin. Genet. 27, 127-133.

235-242.

12, 75-82.

263-277.

pp. 215-228.

MacDonald, W. C., W. 0. Dobbins, I11 & C. E. Rubin (1965). Studies of the familial nature of coeliac sprue using biopsy of the small intestine. N . Engl. J . Med. 272, 448456.

Mann, D. L., S. I. Katz, D. L. Nelson, L. D. Abelson & W. Strober (1976). Specific B-cell antigens associated with gluten-sensitive enteropathy and dermatitis herpetiformis. Lancet i, 11&111.

McKenna, R., F. M. Stevens, B. McNicholl, S. Scholz, E. Albert & C. F. McCarthy (1983). Family and population studies of HLA and coeliac disease in the West of Ireland. Tissue Antigens 22, 175-181.

Mearin, M. L., I. Biemond, A. S. Peiia, I. Polan- co, C. Vazquez, G. Th. M. Schreuder, R. R. P. DeVries & J. J. van Rood (1983). HLA-DR phenotypes in Spanish coeliac children: their contribution to the understanding of the genetics of the disease. Gut 24, 532-537.

Meeuwisse, G. W. (1970). Diagnostic criteria in coeliac disease. Acta Paediatr. Scand. 59, 46 1-463.

Mylotte, M., B. Egan-Mitchell, C. F. McCar- thy & B. McNicholl(l973). Incidence of coeliac disease in the West of Ireland. Br. Med. J .

Mylotte, M., B. Egan-Mitchell, P. F. Fottrell, B. McNicholl & C. F. McCarthy (1974). Family studies in coeliac disease. Q. J. Med. 43,

O’Farrelly, C., J. Kelly, W. Hekkens, B. Bradley, A. Thompson, C. Feighery & D. G . Weir (1983). Alpha gliadin antibody levels: a sero- logical test for coeliac disease. Br. Med. J . 286,

Peiia, A. S., D. L. Mann, N. E. Hague, J. A. Heck, A. van Leeuwen, J. J. van Rood & W. Strober (1978). Genetic basis of gluten-sensitive enteropathy. Gastroenterology 75,23&235.

Peiia, A. S., I. Biemond, P. C. M. Rosekrans, A. van Leeuwen, I. Schreuder & J. J. van Rood (198 I ) . DR locus-controlled B-cell alloantigens in coeliac disease in The Netherlands. In Gen- etics of Coeliac Disease. R. B. McConnell (ed). Lancaster, MTP Press Ltd. pp. 161-168.

Polanco, I., I. Biemond, A. van Leeuwen, I. Schreuder, P. Meera Khan, J. Guerrero, J. D’Amaro, C. Vazquez, J. J. Van Rood & A. S. Peiia (1981). Gluten sensitive enteropathy in Spain: genetic and environmental factors. In Genetics of Coeliac Disease. R. B. McConnell (ed). Lancaster, MTP Press, Ltd. pp. 211-231.

Robinson, D. C., A. J. Watson, E. H. Wyatt, J. M. Marks & D. F. Roberts (1971). Incidence of small-intestinal mucosal abnormalities and

1, 703-705.

359-369.

2007-201 0.

198 L I N E T A L .

of clinical coeliac disease in the relatives of children with coeliac disease. Gut 12, 789-793.

Robinson, B. N., D. F. Roberts, B. A. Mather, R. Nelson & A. S. Rowlatt (1980). Coeliac disease and HLA: a family study. 1. Imrnuno- genet. 7 , 381-391.

Rolles, C. J., T. B. Kyaw-Myint, W.-K. Sin & C. M. Anderson (1981). Familial incidence of asymptomatic coeliac disease. In Genetics of Coeliac Disease R. B. McConnell (ed). Lanca- ster, MTP Press Ltd., pp. 235-243.

Rotter, J. I. & E. M. Landaw (1984). Measuring the genetic contribution of a single locus to a multilocus disease. Clin. Genet. 26, 529-542.

Shipman, R. T., A. L. Williams, R. Kay & R. R. W. Townley (1975). A family study of coeliac disease. Aust. N . 2. J. Med. 5 , 25&255.

Stokes, P. L., P. Asquith & W. T. Cooke (1973). Genetics of coeliac disease. Clin. Gastroenterol.

Stokes, P. L., P. Asquith, G. K. T. Holmes, P. MacKintosh & W. T. Cooke (1972). Histocom- patibility antigens associated with adult coeliac disease. Lancet ii, 162-164.

Stokes, P. L., R. Ferguson, G. K. T. Holmes &

2, 547-556.

W. T. Cooke (1976). Familial aspects of coeliac disease. Q. J . Med. 45, 567-582.

Strober, W. (1980). Genetic factors in gluten-sensitive enteropathy. In Genetics and Heterogen- eity of Common Gastrointestinal Disorders. J. I. Rotter, I. M. Samloff & D. L. Rimoin (eds). New York, Academic Press, pp. 243-259.

Tiwari, J. L., H. Betuel, L. Gebuhrer & N. E. Morton (1984). Genetic epidemiology of coeliac disease. Genet. Epidemiol. 1, 3742.

van Rood, J. J., J. P. van Hooff & J. J. Keuning (1 975). Disease predisposition, immune re- sponsiveness and the fine structure of the HL- A supergene. A need for a reappraisal. Trans- plant. Rev. 22, 75-104.

Address: Dr. Jerome I . Rotter Division of Medical Genetics Harbor- UCLA Medical Ctr 1000 West Carson Street Torrance, CA 90509 USA

use of hla marker associations and hla haplotype linkage to estimate disease risks in families with...

Documents