DNA RESTRICTION FRAGMENT LENGTH POLYMORPHISMS AS MARKERS OF FAMILIAL CORONARY HEART DISEASE

1407

and cephalexin. These latter seem first observations contradictory to our previous studied individually the same infants in findings we propose that the However, observations). (unpublished beneficial ecological impact of increasing local cefuroxime usage is attributable to a concomitant reduction of local ampicillin use. Thus, even though both agents produced resistant strains in the individually treated newborn baby (unpublished data), the ecological impact of ampicillin was so great that the consequences of reducing ampicillin use outweighed that of the increased cefuroxime use. It is noteworthy that the high frequency of p-lactam resistant enteric bacteria associated with an ampicillin-based policy (and the decreased frequency associated with cefuroxime) was much the same, or more pronounced, in babies who had not received antibiotics and in those who had. Although suspected by Borderon et al,9 these similarities have not previously been clearly demonstrated. We have shown that ampicillin use correlated with nosocomial spread of certain multiply p-lactam resistant strains of E coli, Klebsiella spp, and Enterobacter Spp.13 By unknown means ampicillin (but not cefuroxime) accelerated the local spread of certain drug resistant strains of Enterobacteriaceae, thus accounting for an ecological impact among babies not receiving antibiotics. Several workers have suggested that increasing the use of cephalosporins risks the emergence of mutants of Enterobacter with stable derepressed production of chromosomal P-lactamase, which would thus be highly resistant to most p-lactam agents.14-18 In our study the total number of Enterobacter spp in each ward was too low for statistical analysis. Both ampicillin and cefuroxime therapy, however, increased the risk of individual colonisation by Enterobacter spp and the emergence of cefuroxime and cefotaxime resistance in such strains. Nevertheless; ampicillin seemed to promote local spread of resistant Enterobacter (unpublished data). Besides antibiotic use, institutional factors, as well as bacterial and host factors, may determine the microflora of infants in neonatal wards. Findings from one unit-eg, during an outbreak of infection-may reflect a special situation making interpretation of the results doubtful. Prospective multicentre studies like ours should be done before we draw conclusions about the impact of individual antibiotics on bacterial ecology in neonatal units. Our fmdings illustrate two novel aspects of antibiotic ecology-namely, a greater risk of emergence of drug resistance due to a related agent rather than to the drug itself, and a clear indirect impact of an antibiotic on the microflora of untreated patients.

ampicillin, cefuroxime, at

We thank the staff at the participating paediatric departments for their contributions, Mrs Birgitta Berglund for technical assistance, and Dr Christopher Korch for criticism of the script. This study was supported by grant B88-16X-08302-OlA from the Swedish Society for Medical Sciences, the First of May Flower Campaign for Childrens’s Health, the Karolinska Institute, Stiftelsen Allinanna BB:s Minnesfond, and Stiftelsen Samariten.

Correspondence should be addressed to K. T. REFERENCES 1. Selden R, Lee S, Wang WLL, Bennet JV, Eickhoff TC. Nosocomial Klebsiella pneumoniae infections: intestinal colonization as a reservoir. Ann Intern Med 1971; 74: 657-64. 2. Bourillon A, Brackman D, Bousougant Y, de Paillerets F. Cefotaxime effects on the intestinal flora of the newborn Dev Pharmacol Ther 1984; 7 (suppl 1) 144-49. 3. Bryan CS, John JF Jr, Pai S, Austin TL. Gentamicin versus cefotaxime for therapy of neonatal sepsis: relation to drug resistance. Am J Dis Child 1985; 139: 1086-89.

DNA RESTRICTION FRAGMENT LENGTH POLYMORPHISMS AS MARKERS OF FAMILIAL CORONARY HEART DISEASE W. H. PRICE A. H. KITCHIN P. R. S. BURGON

S. W. MORRIS P. R. WENHAM P. M. DONALD

MRC Human Genetics Unit, University Department of Medicine, and Department of Clinical Chemistry, Western General Hospital,

Edinburgh men aged 30 to 59 years had relatives who had had coronary first-degree heart disease (CHD) by the age of 60; men with a family history of CHD were twice as likely to have CHD themselves, compared with those without such a family history. At least 75% of the difference was accounted for by CHD in men with minor alleles of 4 restriction fragment length polymorphisms (RFLPs) in the region of the apolipoprotein (apo) AI and apo CIII genes. The RFLPs were identified with the restriction enzymes XmnI, PstI, MspI, and SacI. Each polymorphism has two alleles (major and minor), designated X1 and X2, P1 and P2, M1 and M2, and S1 and S2, respectively. In men with any one of the minor alleles, a family history of CHD was associated with a 234% increase in CHD prevalence. In men with major alleles only, a family history of CHD was not associated with any significant increase in CHD. The effect of the minor alleles was not significantly altered when plasma lipids and other variables except for age were taken into account. The association between the minor RFLP alleles and polymorphic gene variants (probably the apo AI, apo CIII, or both genes) which enhance liability to CHD accounted for almost 20% of total CHD in this population.

Summary

246 of 713

4. Modi N, Damjanovic V, Cooke RVI. Outbreak of cephalosporin resistant Enterobacter cloacae in a neonatal intensive care unit. Arch Dis Child 1987; 62: 148-51. 5. Tullus K, Fryklund B, Berglund B, Kallenius G, Burman LG. Influence of age on the fecal carriage of P-fimbriated Escherichia colt and other Gram-negative bacteria in hospitalized neonates. J Hospital Infect 1988; 11: 349-56. 6. Tullus K, Berglund B, Fryklund B, Kuhn I, Burman LG. Influence of antibiotic therapy of faecal carriage of P-fimbriated Escherichia colt and other Gram-negative bacteria in neonates. J Antimicrob Chemother 1988; 22: 563-68. 7. Kühn I. Biochemical fingerprinting of E coli: a simple method for epidemiological investigations. J Microbial Methods 1959; 3: 159-70. 8. Alestig K, Bertler Å, Brorson JE, et al. A revised system for antibiotic sensitivity testing. Scand J Infect Dis 1981; 13: 148-52. 9. Borderon JC, Gold F, Laugier J. Enterobacteria in the neonate. Normal colonization and antibiotic-induced selection. Biol Neonate 1981; 39: 1-7. 10. Lincoln K, Lidin-Janson G, Winberg J. Resistant urinary infections resulting from changes in resistance pattern of faecal flora induced by sulphonamide and hospital environment. Br Med J 1970; 3: 305-09. 11. Sturtevant AB Jr, Cassel GH1, Bobo RA, Feary TW. Effect of antibiotic treatment on the incidence of infectious drug resistance among intestinal lactose-fermenting bacteria isolated from burn patients. Infect Immun 1971; 3: 411-15. 12. Pollack M, Charache P, Nieman RE, Jett MP, Reinhardt JA, Hardy PH Jr. Factors influencing colonisation and antibiotic-resistance patterns of gram-negative bacteria in hospital patients. Lancet 1972; ii: 668-71. 13. Tullus K, Berglund B, Fryklund B, Kühn I, Burman LG. Epidemiology of fecal strains of the family Enterobacteriaceae in 22 neonatal wards and influence of antibiotic policy. J Clin Microbiol 1988; 26: 1166-70. 14. Weinstein RA. Endemic emergence of cephalosporin-resistant Enterobacter: relation to pnor therapy. Infect Control 1986; 7: 120-23. 15. Prevot MH, Andremont A, Sancho-Garnier H, Tancrede C. Epidemiology of intestinal colonization by members of the family Enterobacteriaceae resistant to cefotaxime in a hematology-oncology unit. Antimicrob Agents Chemother 1986; 30: 945-47 16. Sanders CC, Sanders WE Jr. Clinical importance of inducible beta-lactamases in Gram-negative bacteria. Eur J Clin Microbiol 1987; 6: 435-37 17. Wiedemann B. Selection of beta-lactamase producers during cephalosporin and penicillin therapy. Scand J Infect Dis 1986; 49 (suppl): 100-05. 18. Livermore DM. Clinical significance of beta-lactamase induction and stable derepression in Gram-negative rods. Eur J Clin Microbiol 1987; 6: 439-45.

1408 TABLE I—CHARACTHRfS’nCS OF STUDY POPULATION

Introduction EPIDEMIOLOGICAL studies indicate that genetic factors influence an individual’s liability to coronary heart disease (eHD).1.2 This effect is due partly to the genetic influence on conditions that predispose to CHD-such as diabetes mellitus, hypertension, and hyperlipidaemia-and partly to a genetic contribution independent of other risk factors. The frequency of the disease in close relatives of people who have CHD is consistent with a polygenic mode of inheritance. The genes responsible may belong to normal polymorphisms.3 In 1983, Karathanasis and colleagues4 described a rare structural DNA rearrangement that involved the apolipoprotein (apo) AI and Ciii genes in sisters who suffered from early atherosclerosis and CHD. Apo AI and apo Ciii are the major apoproteins of high density and very low density lipoproteins, respectively; the genes lie 2-5 kb apart on the long arms of chromosome 11. Eleven DNA restriction fragment length polymorphisms (RFLPs) have been identified in the apo AI/CIII gene region and associations with hyperlipidaemia, CHD, or both conditions have been described for at least 6;5 however, other studies6,7 have not confirmed the disease association of these RFLPs, possibly because of the high prevalence of non-familial CHD in some populations. We have compared the prevalence of CHD in relation to apo Al and apo Cm RFLP genotypes in those with and without a family history of CHD and have also examined the relation of these polymorphisms to plasma cholesterol and

I

I

Results shown as number (% of age-group). No = number in age-group; CHD = coronary heart disease; HT = hypertension ; BMI=body mass index (weight/height); TC=plasma total cholesterol; TG plasma triglyceride; FH positive family history of CHD in a first-degree relative before age 60 (details of FH not known for 23 =

=

subjects). TABLE II-MEANS OF DIAGNOSIS OF CHD

*Minnesota codes 1 1, 1-2, and 7 1. tMinnesota codes 13, 41,42, 4 3, 5 1, 52, and 5-3. tWHO chest pain questionnaire for history of angina or possible myocardial infarction.

triglyceride. Subjects and Methods The subjects were men aged 30 to 59 years who were participants in a CHD prevention programme of a general practice in the Leith district of Edinburgh. All 1059 eligible men had been invited to participate in a letter from their general practitioner. 713 attended a clinic where WHO cardiovascular and smoking questionnaires8 and a Scottish Heart Health Study family history questionnaire (Tunstall-Pedoe H, Smith C, personal communication) were completed. Investigation included measurements of height; weight; blood pressure; plasma total cholesterol, HDL-cholesterol, and triglyceride; and a 12-lead electrocardiograph. If plasma total cholesterol or triglyceride exceeded 6-4 mmol/1 or 22 mmol/l, respectively, the measurements were repeated on blood samples obtained after an overnight fast. 15 ml of blood were collected into mucous heparin for DNA studies and DNA was extracted by the method of Kunkel et al.9 Aliquots of DNA were digested with the enzymes Xmni, Mspi, Psti, and SacI according to the manufacturers’ recommendations. DNA fragments were separated by gel electrophoresis and after denaturation they were transferred onto nitrocellulose filters by Southern blotting. Filters were hybridised with one of two 32P labelled probes: a 2-2 kb tandem repeat of the full length apo Ai gene, cloned in pSV2 (kindly provided by Prof J. L. Breslow) and a cDNA apo AI clone (pBAl ; kindly provided by Dr F. E. Barelle and Dr C. C. Shoulders). The filters were then autoradiographed. Plasma total cholesterol was measured by a cholesterol oxidase method, HDL-cholesterol by phosphotungstate precipitation, and triglyceride by a glycerol kinase method. Frequencies were compared by X2 analysis or the exact probability test when numbers were small. The independent contributions of multiple variables were assessed simultaneously by

of 100 non-respondents were compared with a sample of 100 who did attend: there were no significant differences in social class or in known personal history of CHD or hypertension, but significantly more non-respondents were cigarette smokers. Baseline characteristics of the study population are shown in table 1. Just over one third of the men examined had a family history of CHD, defmed in this study as a history of CHD in a first-degree relative before the age of 60 years. The affected relative was the father in 52%, mother in 22%, brother in 26-5%, and sister in 11%, 16% had a family history of CHD in more than one relative, and 4% in more than two; both parents were affected in 2%. 51% of the men had electrocardiographic evidence of CHD, as defined by the Minnesota codeand 7-5% had a history of angina or possible myocardial infarction. The overall prevalence of CHD was 11’2% (table 11), the prevalence of CHD in men with a positive family history of CHD was significantly higher than in men without such a men

family history (15-9% vs 8-1%, p < 0-005). Autoradiographs of DNA from subjects homozygous and heterozygous for major and minor alleles of the .X?KMl, Mspt, Pstl, and Sacl RFLPs are shown in the figure. Major alleles, designated X 1, Ml, Pl, and Swere identified by fragment band sizes of 8-3 kb, 1 -0 kb, 2-2 kb, and 4-2 kb, respectively. TABLE III-ALLELE AND GENOTYPE

FREQUENCIES IN TOTAL

POPULATION

logistic regression analysis. Results

Completed questionnaires, electrocardiographs, and were obtained from 713 men (67% of those for inclusion), although HDL-cholesterol was eligible in measured only 538. Medical records of a random sample blood samples

*Results shown

as

observed

Weinberg equation).

frequency (expected frequency from Hardy

1409 TABLE V- DISTRIBUTION AND RELATIVE PREVALENCE OF CHD BY FAMILY HISTORY

AND APO AI/CIII RFLP ALLELE STATUS

Values shown as number (%). HM=homozygous; HZ=heterozygous; NS not significant. *Prevalence of M2 and S2 alleles combined because of tight linkage disequilibrium between these two alleles. =

and Autoradiographs of DNA from subjects homozygous alleles at polymorphic and major minor for heterozygous restriction enzyme sites identified by (a) Xmni, (b) Mspl, (c) PstI, and (d) Saci.

MspI and Sacl digests were hybridised with pBAl probe; XmnI and PstI with pSV2 apo Ai probe. DNA fragment band sizes in kilobases.

Minor alleles X2, M2, P2, and S2, were identified by band sizes of 6-6 kb, 1.7 kb, 3-2 kb, and 3-2 kb, respectively. The population frequencies of the minor alleles and the genotype frequencies for each RFLP are shown in table III: distribution of genotype frequencies did not depart significantly from Hardy Weinberg predictions. Allele frequencies in subjects with CHD, a positive family history, or raised plasma total cholesterol or triglyceride are shown in table iv: no significant differences from total population frequencies were observed. An analysis of estimated pairwise haplotype frequencies for evidence of linkage disequilibrium (D) as described by Thompson et aPO showed a significant positive linkage disequilibrium between the M2 and S2 alleles (D= 66%; p < 0-001), but not for any other allelic pairs. The haplotype frequencies (shown as haplotype, number [%]) were: X1M1P1S1, 971 (7345); X2M1P1S1, 134 (12-4); X1M2P1S2, 75 (567);

X1M1P2S1,74 (559); XlMlPlS2,23 (1-74); X1M2P1S1, 13 (0-98); X1M2P2S1, 2 (0-15). Full data on family history of CHD and genotypes of all 4 RFLPs were obtained for 667 men, of whom 71 (10-6%) had a personal history or electrocardiographic evidence of CHD. Table v shows the distribution of CHD by apo Al/Clll RFLP genotype in men with and without a family history of CHD. In men with a family history, the prevalence of CHD was significantly higher in those whose

genotype included a minor allele of one of the 4 RFLPs, but not in those who had RFLP major alleles only. If an expected frequency of 0-079 due to non-familial factors (the frequency of the disease in men without a family history of CHD) is assumed, the proportion of the familial component attributable to an association with minor alleles is the number of men with CHD who have a positive family history (FH) and minor alleles in excess of the expected, divided by the total number of FH-positive men with CHD in excess of expected, and equals 0-759 (from table v: 22 - [103 x 0-079]/37 - [237 x 0079]). The number of FHpositive men with CHD and minor alleles in excess of the expected, expressed as a proportion of the total number of men with CHD (13-86/71 = 0-195) is a measure of the total CHD load attributable to the association with the minor alleles. 28-6% of men had plasma total cholesterol above 6-4 mmol/1 and 244% had plasma triglyceride above 22 mmol/1. In men with a family history of CHD these levels were exceeded in 35-8% and 31-3%, respectively (p = 0-025 and 0-001, respectively, compared with men without a family history of CHD). The prevalence of CHD in men with raised plasma total cholesterol (9-8%) and triglyceride (10-6%) were not significantly different from that in the total The increased frequency of raised plasma cholesterol in FH-positive men was not significantly influenced by apo Al/Clll RFLP status, but the increased frequency of raised plasma triglyceride in FH-positive men

population.

TABLE VI-PREVALENCE OF HYPERLIPIDAEMIA BY APO AI/CIII

GENOTYPE

TABLE IV-MINOR ALLELE FREQUENCIES

Abbreviations

as

for table

i.

Values shown as number (%). Abbreviations as for table V *Not significant; tp < 0-01; tp < 0-025; Up < 0-0005.

1410 TABLE VII-DISTRIBUTION OF CHD IN HYPERLIPIDAEMIC MEN BY APO AI/CIII GENOTYPE

CHD values shown as number

(%). NS

=

not

significant.

occurred only in those with minor RFLP alleles (table VI). Mean plasma HDL-cholesterol values were not significantly different in any of the groups examined. The raised CHD prevalence in FH-positive men with apo AI/CIII RFLP minor alleles was independent of raised plasma total cholesterol or triglyceride (table VII) and of variation in HDL-cholesterol. Logistic regression analysis to assess simultaneously the independent contributions to CHD of a positive family history and the presence of a minor allele showed no significantly increased risk for either factor independently, but the combination of both factors gave a relative odds value of 2 65 (95% confidence limits 1 007-11). When the other risk factors of age, hyperlipidaemia, smoking, and hypertension were included the relative odds value was 2-50 (0-92-6-76), of these other factors, only age was independently significant with a relative odds value of 1-24 (1.05-1.46).

Discussion The prevalence of CHD observed in this study was 11-2% for all ages and 12 6% in men aged 40 to 59 years. This prevalence is lower than has been recorded in other British studies.11-13 Mild and severe angina were diagnosed in our study as frequently as in these other studies, but there were fewer questionnaire diagnoses of myocardial infarction. Electrocardiographic evidence of ischaemia was found with the same frequency as in the Whitehallll and UK Heart Disease Prevention Project12 studies, but was less common than in the British Regional Heart Study13possibly because the BRHS used Macfarlane’s orthogonal electrocardiograph lead system with computerised analysis and interpretation whereas we and the other 2 studies used the Minnesota coding of the conventional electrocardiograph. When we applied the suggested Minnesota coding equivalents of the orthogonal system13 to our data the CHD frequency was not increased, which indicates that the criteria used by the computerised system are wider than the Minnesota code equivalents. Other possible explanations for the discrepancies are a lower CHD incidence in Edinburgh or an unconscious bias excluding CHD patients in our sample-a possibility supported by the higher frequency of cigarette smokers in non-respondents. RFLP status was determined without knowledge of the clinical diagnosis, so those with and without minor alleles should have been affected equally by any incomplete ascertainment of CHD. As in most recent studies of family history in CHD, we have found that in men aged 30 to 59 years the prevalence of CHD is almost doubled in those with affected first-degree relatives.2 In an earlier study6 we were unable to

demonstrate an association between an apo AI/CIII RFLP and myocardial infarction in men under the age of 50 years. Similarly, in the present survey, we found no association between Al/Clll RFLPs and CHD in general. However, we found a significant increase in disease prevalence in FH-positive men with each of the minor RFLP alleles studied. This increase accounted for more than 75% of the greater familial prevalence of the disease, and for almost 20% of CHD in the population. In subjects with major alleles only, prevalence of CHD was not significantly affected by a family history of the disease, and in men with no affected first-degree relatives RFLP alleles made no difference to the prevalence of the disease. Most polymorphic DNA sites are outside the genes or in non-coding or untranslated regions within a gene. They are therefore unlikely to be responsible for altered phenotypic expression of the gene. The observed association is most probably a result of linkage disequilibrium (cosegregation of two genes of genomic sequences with a greater frequency than expected) between the RFLPs and polymorphic genes which contribute to the variance in genetic susceptibility to CHD. We suggest that each minor allele is linked to a genetic variant that enhances liability to CHD: since linkage disequilibrium occurs between genetic loci that are in close proximity, the genes most likely to be implicated are the apo AI and apo CIII genes. The same minor apo AI/ClII alleles are associated with triglyceride values above 2 -2 mmol/1, but again only in those with a positive family history of CHD. The prevalence of CHD in FH-positive men with minor RFLP alleles is significantly raised in those with hypertriglyceridaemia, but even more so in those with triglyceride levels considered to be within the normal range. There is no significant association of hypercholesterolaemia or of reduced HDLcholesterol levels with minor RFLP alleles in FH-positive men. (The increased familial prevalence of CHD in those with minor alleles is evident only in men with cholesterol values below 6-5 mmol/1). The raised prevalence of CHD in men with minor RFLP alleles therefore appears to be independent of variation in plasma lipid levels although a change in lipoprotein structure or an alteration in apolipoprotein AI and Ciii regulation, which may influence liability to atherosclerosis and CHD, are not excluded. The significant effect of age could be accounted for by the greater proportion of older men with affected first-degree relatives under 60 years of age. We thank Prof J. L. Breslow, Dr F. E. Barelle, and Dr C. C. Shoulders for the apo Ai probes, Dr A. G. Donald and partners for access to their patients, and Dr R. A. Elton for statistical advice. This work was supported by the MacKenzie trust and the Margaret Kennedy fund.

Correspondence should be addressed to W. H. P., MRC Human Genetics Unit, Western General Hospital, Crewe Road, Edinburgh EH4 2XU. REFERENCES 1. Neufeld HN, Goldbourt U. Coronary heart disease: genetic aspects. Circulation 1983, 67: 943-54. 2. Goldbourt U, Neufeld HN. Generic aspects of arteriosclerosis. Arteriosclerosis 1986, 6: 357-77. 3. Berg K. Genetics of coronary heart disease. Prog Med Genet 1983; 5: 35-90 4. Karathanasis SK, Norum RA, Zannis VI, Breslow JL. An inherited polymorphismm the human apolipoprotein AI gene locus related to the development of atherosclerosis. Nature 1983; 301: 718-20. 5. Breslow JL. Apolipoprotein genetic variation and human disease. Physiol Rev 1988; 68: 85-132. 6 Morris SW, Price WH. DNA sequence polymorphisms in the apolipoprotein A-I/C-III gene cluster. Lancet 1985; ii: 1127-28. 7 Paulweber P, Fnedl W, Krempler F, Humphries SE, Sandhofer F. Genetic variation m the apolipoprotem AI-CIII-AIV gene cluster and coronary heart disease. Atherosclerosis 1988; 73: 125-33.

ferences con

Re

tinuted

at

foot

of next

page

1411

SEROLOGICAL DIAGNOSIS OF SALMONELLA INFECTIONS BY ENZYME IMMUNOASSAY RISTO VUENTO OUTI ISOMAKI KAISA GRANFORS Department of Medical Microbiology, Turku University, Turku, Finland

An enzyme immunoassay (EIA) for the detection and measurement of serum IgM, IgG, and IgA antibodies to salmonella was developed with commercially available lipopolysaccharides (LPSs) of Salmonella typhimurium and S enteritidis combined as antigen. Of 130 sera from patients with culture-confirmed salmonella infections, 115 (88·5%) were positive in this assay. The classic Widal agglutination test was positive in only 50 (38·5%) cases. This EIA method offers a substantial advance in the serological diagnosis of acute salmonella infections; it detects antibodies to the salmonellae of groups B and D, which constitute about 70% of culture-positive cases of human salmonellosis. Antibodies to other salmonellae are also detected. This EIA is particularly valuable for the detection of salmonella antibodies during post-infectious complications when isolation of the organism is often no longer possible.

yersinia infections :24-27 this has been especially valuable in the retrospective diagnosis of yersinosis as a cause of reactive complications25-27 because of its better sensitivity for detection of IgG and IgA than the agglutination method. In this report we describe a similar method for the routine serological diagnosis of salmonella infections with commercially available lipopolysaccharides (LPSs) of S typhimurium and S enteritidis as antigen.

Summary

Introduction THE mainstay in serological diagnosis of salmonella infections is still the Widal agglutination test,1 although the value of this method is often questioned.2,3Another test is the enzyme immunoassay (EIA) by which concentrations of antibodies against salmonellae in human sera can be measured:4-11 serotype-specific EIA, for example, has proved useful in the detection of antibody responses to Salmonella typhimurium during an epidemic.6,10,11 However, the value of EIA for routine diagnostic purposes is limited because there are numerous salmonella serotypes and antigens. Thus, neither EIA nor Widal agglutination is of great value for routine diagnostic testing. Of the salmonellae that cause infections in man, about 70%, including S typhimurium and S enteritidis, carry identical 0 antigens (groups B and D).7,12-16 The same serotypes, S typhimurium, and occasionally S enteritidis have been associated with reactive arthritis as a post-infectious complication, especially in HLA-B27 positive individuals.10,11,17-23 By the time arthritic complications appear, stool cultures are often negative. Thus, laboratory diagnosis must rely on the detection of specific antibodies in the serum. The separate measurement of yersinia-specific IgM, IgG, and IgA responses has offered a great advantage over classical agglutination methods in the serodiagnosis of W. H. PRICE AND OTHERS:

REFERENCES—continued

8. Rose GA, Blackburn H, Gillum RF, Prineas RJ. Cardiovascular survey methods, 2nd ed. Geneva: World Health Organisation, 1982. 9. Kunkel LM, Smith KD, Boyer SH, et al. Analysis of Y chromosome specific reiterated DNA in chromosome variants. Proc Natl Acad Sci USA 1977; 74: 1245-49. 10. Thompson EA, Deeb S, Walker D, Motulsky AG. The detection of linkage disequilibrium between closely linked markers: RFLPs at the AI-CIII apolipoprotein genes. Am J Hum Genet 1988; 42: 113-24. 11. Reid DD, Brett GZ, Hamilton PJS, Jarrett RJ, Keen H, Rose G. Cardiorespiratory disease and diabetes among middle-aged male civil servants. A study of screening and intervention. Lancet 1974; i: 469-73. 12. WHO European collaborative group. Multifactorial trial in the prevention of coronary heart disease. 1: recruitment and critical findings. Eur HeartJ 1980; 1: 73-80. 13. Shaper AG, Cook DG, Walker M, Macfarlane PW. Prevalence of ischaemic heart disease in middle aged British men. Br Heart J 1984; 51: 595-605.

Patients and Methods Patients Serum samples from 130 patients with salmonella infection, confirmed by positive stool culture, were studied. 88 patients with S typhimurium infection were from four localities: 63 from the outbreak in Mellilall (close to Turku) in 1982; 9 from the outbreak in Turku in 1986; 13 from the infectious disease unit of the City Hospital in Turku; and 3 from Turku Health Centre. Sera from 20 patients with S enteritidis infection and from 22 patients infected with other salmonella serotypes-S infantis (6); S virchow (4); S emek (2); S stanley (2); and 1 each with S blockley, S goldcoast, S münchen, S anatum, S panamn, S san diego, S bovismorbificans, and Salmonella group D-were from the Turku Health Centre. Sera from the patients with S typhimurium infection from the outbreak in Mellila were taken at 2-5 months, whereas all the other sera were collected 1-3 weeks after infection. The 9 patients from the S typhimurium outbreak in Turku were followed up for 7-11 months; 53 serum samples were available. Otherwise, 1 sample per patient was studied. Ages of the patients ranged from 1 to 76 years (13 children, 117 adults).

Controls Two groups of control sera were used. Firstly, sera from 100 healthy blood donors from the same area as the patients were tested. Secondly, to establish the specificity of the test, sera from patients with other acute infections were examined: 35 had Yersinia enterocolitica 0:3 infection; 4 had shigella infection (3 Shigella sonnei, 1 S flexnerii); 3 had Campylobacterjejuni infection; 7 had Chlamydia trachomatis infection; and 19 had viral infections (5 respiratory syncytial virus, 2 influenza A virus, 8 influenza B virus, 2 measles virus). The laboratory diagnoses of yersinia infections were made by positive stool culture or high concentrations of serum antibodies (EIA); of shigella and campylobacter infections by positive stool culture; and of chlamydial and viral infections by high concentrations of serum antibodies (EIA). Samples from patients with yersinia, shigella, and campylobacter infections were from the diagnostic laboratory of the Department of Medical Microbiology, and samples from patients with chlamydial and viral infections from the Department of Virology, Turku University.

Salmonella Antigens for EIA

Phenol/water-extracted LPSs of S typhimurium and S enteritidis (Sigma, Montana, USA) were used either individually (single LPS EIA) or together (pooled LPS EIA) as antigen. Sodium dodecyl sulphate (SDS) extracts of whole salmonella bacteria were also tested as antigen. The S typhimurium strain (ATCC 1311) was obtained from the American Type Culture Collection (Rockville, Maryland, USA). The S enteritidis strain was a clinical isolate from the Department of Medical Microbiology. To prepare antigen, 10 ml of an overnight nutrient broth culture of S typhimurium or S enteritidis was diluted 1 in 40 in fresh nutrient broth and grown on a shaker for 4 h at 37°C. After three washings, the bacteria were treated with 0.1 % (weight/volume) SD S for 1 h at 37°C to obtain an antigen extract (SDS-extract).

EIA for IgM, IgG, and IgA Salmonella Antibodies The EIA procedure was similar to that described for detection of anti-yersinia antibodies.24 Polystyrene microtitre plates (Nunc, Roskilde, Denmark) were coated with bacterial antigens (100