Ancient HLA: a preliminary survey

Evison, M.P., Fieller, N.R.J. and Smillie, D.M. (1999). Ancient HLA: a preliminary survey. Ancient Biomolecules, 3(1), 1-28. ISSN 1358-6122

Ancient Biomolecules, Vol. 3, pp. 1-28 Reprints available directly from the publisher Photocopying permitted by license only

J) 1999 OPA (Overseas Publishers AssociatIOn) N.V

Published by license under

the Harwood Academic Publishers imprint.

part of The Gordon and Breach Publishing Group.

Printed in MJIClysia.

Ancient H LA: A Preliminary Survey MARTIN P. EVISON a ,*, NICHOLAS RJ FIELLERb and DAVID M. SMILLIE' a Department

of Forensic Pathology, Umversity of Sheffi eld, The Medlco- U gal Centre, Watery Sireet, Sheffield S3 7ES, UK;

bSchoolof Mathematics and Stat/sties, Umverslty of Sheffield, Sheffield 53 7RH, UK;

'Trent Centre, National Blood Service, Longlel) Lane, Sheffield 55 7iN, UK

(Received 1 May 1998; In final form 24 October 1998)

Recovery of single copy nuclear DNA sequences from ordinary archaeological tooth and bone speci­ mens is necessary for testing models of the history and evolution of nuclear genes, and to complement results from ancient mitochondrial DNA analyses. Here we present a preliminary survey of ancient single copy nuclear DNA survival in archaeological material originating from a number of sites in Europe. The HLA complex and X-Y homologous amelogenin gene were chosen as targets because of their value in the study of kinship, gender, demo­ graphy and immunity, and for the potential for DNA-independent verification of results. HLA-B27 and amelogenin DNA analysis of modem blood sam­ ples (n = 431) were used to establish the reliability of the protocols. DNA purification methods were refined using forensic material as a model substrate, and applied to a study of amelogenin and HLA-DPBl survival in forensic (n = 82) and archaeological (n = 92) specimens. Generalised linear interactive modelling established that there was no relationship between positive results and the presence or level of contamination detected in negative controls, Most promising results in this preliminary survey were obtained from specimens exhibiting superior gross morphological preservation from archaeological sites in Greece. Amplification refractory mutation system polymerase chain reaction analysis of Palaeolithic bone from Thessaly yielded an identical HLA-DRBl type from samples obtained from separate specimens from the same skeleton.

* Corresponding

Keyiuords: Amelogenin, Ancient DNA, HLA, HLA-B27, HLA-D, Mitochondrial DNA

INTRODUCTION The human leukocyte antigen (HLA) system is the most polymorphic genetic system known in humans. It is the human equivalent of the major histocompatibility complex (MHC) and some comparable system is believed to exist in all vertebrates, Allelic variants of HLA are known to be associated with disease and resis­ tance to disease, and a large body of data on HLA diversity in the modern population has been accumulated (Tiwari and Terasaki, 1985; Tsuji et ai., 1992; Charron, 1997). When combined with DNA sexing techniques 'ancient HLA' potentially offers a unique means of investigat­ ing biological relationships in burials, past pop­ ulation movements, the evolution of disease and immunity, and of evaluating phylogenetic models of evolution. In order to assess whether HLA sequences can be reliably recovered from archaeological

author. Tel.: +441142738721. Fax: +44 114 279 8942. E-mail: [email protected].

Evison, M.P., Fieller, N.R.J. and Smillie, D.M. (1999). Ancient HLA: a preliminary survey. Ancient Biomolecules, 3(1), 1-28. ISSN 1358-6122

2

M.P EVISON et al.

material, we conducted a preliminary survey of HLA survival in forensic and archaeological specimens, complemented by DNA sexing using the amelogenin .method and an analysis of a sequence of mitochondrial DNA (mtDNA) in­ cluded to provide a.comparison of mitochondrial and nuclear DNA survival. The results are pres­ ented in this paper. The effectiveness of the DNA analyses used was initially established in a trial utilising a large sample of modern specimens derived from whole peripheral blood. A refinement of the 'silica ' method was used to extract DNA from forensic specimens with a variety of post-mortem histories and then ap plied to a study of human archaeo logical teeth and bone, derived from a number of geographical locations and burial contexts in Europe, varying in age from the Upper Palaeoli thic (circa 14,000 Be) to La te Medieval (circa 1540 AD). Sequence specific primer peR (PCR-SSP) was used to provide a provisional assessment of HLA and amelogenin survival. Sequences of HLA- 0 P B1 were targeted, due the extensive polymorphism at this locus, and HLA-B27, be­ ca use of the associa tion of this allele with ankylos ing spondylitis, a pathology which can be readily detected in archaeological material. Skeletal sex and the ankylosing spondylitis pathology thus offered independent means of verifying ancient DNA results. In addition, potential phylogenetic controls (Paabo, 1989; Handt et ai, 1994; Richards et al., 1995) are offered via polymorphism at the HLA-DPBl locus. Ex­ tensi ve controls were included during extraction, purifica bon and amplification phases. Putative positive results were obtained at a level w hich would make genetic analysis of archaeological specimens viable, but consider­ able contamination was detected. A statistical analysis precluded contamination as a general explanation for the positive resu lts, however. The DNA sequences of a proportion of positive specimens were resolved by dot blotting and amplification refractory muta tion system peR

(peR-ARMS). Explanations for the pattern of sequences observed are discussed in relation to other ancient human nuclear DNA publications and possibilities for future work are considered .

RESULTS Modern Samples Typical results of HLA-B27 and amelogenin PCR analysis are given in Figs. 1 and 2, respectively All HLA-B27 DNA results corresponded to the known serotype at the first or, rarely, the second attempt. Six specimens did not sex correctly and remained incorrect even on retesting; these were one 'false female' and five 'false males'. The frequency of the HLA-B27 phenotype in males and females in the modern sample is given I 2 3 4 5 6 7 8 910 - 796 bp - 135 bp FIGURE 1 Photograph of 1.5% agarose gel illustrating typ­ ica l HLA -B27 results from blood specimens. Normally, the 796 bp C3 / C5 target is present in all samples, but the 135 bp band is present onl y in HLA -B27 positive sa mples. Lanes: 1, HLA-B27 positive control; 2, HLA-B27 negative co ntrol; 3-9, test specimens 273291, 279902, 281205, 281222, 268926, 280661, 280665 respectively; 10, sterile water con­ trot. Test specimens in lanes 3 and 5 are HLA-B2 7 positi ve

1 2 3 4 5 6 7 8 91011 1213 141516

.Wfl!l~~. = 112 bp

106 bp

FIGURE 2 Photograp h of 5% agarose gel illustrating typi ca l amelogenin results from blood specimens. Nor­ mally, two bands are generated for a male (112 and 106 bp) and one for a female 006 bp on ly). Lanes: 1, male positive control; 2, female positive control; 3-16, test specimens

374807, 33511 2, 340123, 340124, 340181, 335520, 335521, 341152, 342225, 343751, 344335, 345049, 345144, 342440 respecti ve ly. Test specimens in lanes 4, 6, 8, etc. are typed as male, those in lanes 3, 5, 7, etc. typed as female .

Evison, M.P., Fieller, N.R.J. and Smillie, D.M. (1999). Ancient HLA: a preliminary survey. Ancient Biomolecules, 3(1), 1-28. ISSN 1358-6122

ANCIENT HLA : A PRELIMINARY SURVEY TABLE I Sex presented

Frequency of HLA-B27 in the sample of modem blood specimens HLA-B27 negative

HLA-B27 positive

%

n

%

n

%

Female Male

9 26

47 10.9

184 212

95.3 891

193 238

44.8 55.2

Total

35

8.1

396

91.9

431

100.0

= total number of specimens.

2 ;< 2 conti.n gency table for HLA-B27 X2 analysis

Sex presenl ed

Sample totals

n

/1

TA BLE II

3

HLA -B2 7 pos itive

HLA-B27 negative

Total

Females Males

9 26

184 212

193 238

Total

35

396

431

1 2 3 4 5 6 7 8 9 10 E B P F 112 bp

in Table L The ratio of HLA-B27 in males to HLA­ B27 in females in the sample is 2.9: 1 (P = 0.74, n = 35). The 2 x 2 contingency table for X2 analy­ sis of the results is given in Table II. The null hypothesis that HLA-B27 phenotype is indepen­ dent of the sex of the donor is rejected (X 2 = 5.60, df=l, level of significance = 0.02). A 95% confi­ dence interval (CI 9s ) for the proportion (P) of males in the HLA-B27 positive population can be calculated from this sample as CI95 = 0.59-0.89 · (Gardner and Altman, 1986). Thus the proportion of HLA-B27 positive males in the sample is 074, and it can be stated with 95% confidence tha t the proportion in the population lies within the range 0.59-0.89.

Forensic Samples The amelogenin PCR yields two bands, corre­ sponding to X and Y, for a male, but only the X band for a female (Fig. 3). Single bands are generated by the amplification of HLA-DPBl sequences (Fig. 4) and sequences of the mito­ chondrial non-coding region V (Fig. 5) To give an estimate of the amount of target present rela­ tive to modern template, the intensity of region V

= 106 bp

FIGURE 3 Photograph of 5% agarose ge l showing typical results for amelogenin analysis of forensic specimens. Lanes: 1-10, tes t specimens F31, F32, F33, F34, F35, F36, F37, F38, F39, F40 respectively; E, extractio n blank; B, silica binding (p urifi ca tion) blank; P, PCR blank; F, PCR positive control DNA (female, 10 ng Ill-I). Faint contamination ca n be distinguished in blanks E and P (see Discussion).

1

2

3

4

5

6

7

8

9 10

E

B

P +

327 bp-

FIGURE 4 Photograph of 1.5 % agarose gel sho wing typi­ cal results for HLA-DPBl analysis of forensic specimens . Single bands are generated by the amplification of the 372bp target sequence. Lanes: 1-10, test specimens F31, F32, F33, F34, F35, F36, F37, F38, F39, F40 respectively; E, extraction blank; B, silica binding (purification) blank; p, PCR blank; +, PCR positive control DNA (lOnglll- 1) .

mtDNA sequences amplified was subjectively rated on a 3 point scale corresponding to prod­ ucts generated from> 200, 2-200 and < 2 pg 111-1 of modern DNA Concentrations of < 0.2 pg 111-1 do not amplify reliably using our system (Evison, 1996) Results for all 119 specimens analysed are given in Appendix I and are summarised in

Evison, M.P., Fieller, N.R.J. and Smillie, D.M. (1999). Ancient HLA: a preliminary survey. Ancient Biomolecules, 3(1), 1-28. ISSN 1358-6122

4

M.P EVISO N et al. TABLE III

Specimen type

Amelogenin

Exh umed bone Bone cross-sections Dried blood spots Forensic bone FOI('" sic teeth Extrilc ted teeth Post-operative bone Totil l

3

4

Tota l

mtDNA region V

HLA-DPBl

n

+

%

n

+

%

n

+

%

n

+

%

37 20 28 8 2 18 6 119

29 13 28 6 2 13 6 97

78 65 100 75 100 72 100 82

26 16 28 7

6 12 12 4

2

.

23 75 43 57 100 80 100

26 16 28 7 2 5 2

19 15 28 7 2 4 2

73 94 100 100 100 80 100

49

86

77

90

89 52 84 22 6 28 10 291

54 40 68 17 6 21 10 216

61 77 81 77 100 75 100 74

n = total number of specimens;

2

Summary of results of PCR analysis of forensic specimens

5

6

+ 7

2

5 2

2 42

86

= nu mber of specimens tha t gil ve PCR products

8

9 10 E

B

P

+

TABLE IV Summary of am eJogenin sexing results for for­ ensic specimens I ncorrect sex

Correct sex

% Correct

Exh umed bone Dried blood spots Forensic bone Forensic teeth Extracted teeth Pos t-operati ve bone

15 0 3 2 0 0

To tal

20

14 28 3 0 13 6 64

48 100 50 0 100 100 76

Specimen type

FIGURE 5 Photograph of 1.5% aga rose gel showing typi­ cal results for mtDNA region V an alys is of fore ns ic speci­ mens. Single bands are generated by the am plification of the 121 bp target sequence. Lanes: 1- 10, test specimens F31, F3 2, F33, F34, F35, F36, F37, F38, F39, F40 respecti vely; E, extraction blank; B, silica binding (purification) blank; P, PCR blank; +, PCR positive control DNA (10 ng Ill-I). Although sensitive to low DNA subs trate concentrations (~250 fg f,ll - J), the PCR may generate high molecular weight ar tefacts at higher concentrations (lanes 1-3 and + ).

Table III. Of these, 82 specimens were analysed in a final study using the refined method (denoted by T in Appendix 1) and 37 specimens were used in methodological development (see below), sometimes leaving insufficient DNA for all three PCR analyses (these are given as 'not attempted' in Appendix D. Table IV shows the amelogenin DNA sexing results for the 82 specimens ana­ lysed in the final stud y. The sex of the donor was known for 66 specimens, enabling the DNA sexing results to be independently verified and an improvement in results to be detected, as a consequence of the methodological refinement. Archaeological Samples Typ ical PCR results for archaeological samples are shown in Figs. 6-8, for amelogenin, HLA­

E

I

2

3

4

5

6

7

8

9

10 11 12 13 14 15

_ -106 bp

FIGURE 6 Photogra ph of 1.5% agarose gel s howing typi ­ cal results of amelogenin PCR of archaeo logical sa mples. Lanes: E, extraction blank control; 1-15, test specimens MAK P480 T662, MAK P480 T667, MA K CT11 SEC(A) P945, MAK T668 P480, KOR TV PI Tn, KOR TXl T3A, EO T33 E0533 (tooth ), KOR TV PI T11 /2. KOR TXI T3A (tooth), MAK CTllSEC (A) P945, MAK T667 P480 (tooth ), MAK T662 P480 (tooth), EO T33 E0533, MAK T668 P480 (too th), THE 27-0,75 43 respectively.

OPBl and mtDNA region V, respectively. Sum­

mary tables of results refer to the 92 speci­ mens extracted and analysed in duplicate (see Appendix Il). Table V includes all archaeological

Evison, M.P., Fieller, N.R.J. and Smillie, D.M. (1999). Ancient HLA: a preliminary survey. Ancient Biomolecules, 3(1), 1-28. ISSN 1358-6122

At'KIENT HlA: A PRELIMINARY SURVEY

1 2 3 4 5 6 7 8 9 10 E B

pee 327 bp

FIGURE 7 Photograph of 15% agarose gel showing typ i­ ca l results of rJlA-DPB l PCR of archa eo logical sam ples. l anes: 1-1 0, test specimens M.273, M.27~, M.2S0, M.300, Torqua y, BOl018, BOl020, BOl029, BOl046, BOl 049 re­ spectively; E, extrac tio n blank; B, silica bi ndi ng (pu rifica­ tion) blank; P, PCR blank; C, positive controls

EEl 2 3 4 5 6 7 8 9 1011 1213 14

121 bp y

-

-

~\-"

FIGURE 8 Photograph of 1.5% agarose ge l showing typi­ cal results of mtONA region V PCR of archaeological sam­ ples. lanes: 1-14, test specimens MAK N80 T662, MAK P480 T667, MAK CTll SEC (A) P945, MAK T668 P480, KOR TV PI Tll , KOR TXI T3A, EO T33 E0533 (too th), KOR TV PI Tll /2, KOR TXT T3A, MAK CTllSEC(A) P945, MAK T667 P480 (tooth), MAK T662 P480 (tooth), EO T33 E0533, MAK T 668 P480 (tooth) respectively; E, ex traction blank controls.

results, irrespective of th e occurrence of contam­ ination_ Table VI includes on ly samples for w hich there was no contamination in any PCR system for either duplicate. Samples in which either one (+ ) or both (++) of the duplicates yielded PCR products are indicated.

Statistical Analysis In order to investigate wh ether the chance of ob ta in ing a positive result from archaeological or forensic material was in any way related to the presence or absence of contamina tion, or level of contamination, a statis tical analysis was per­ form ed by N.F Tests were performed for each of

5

the three sets of archaeo logical samples (regions) and for two sets of forensic samples (blood and bone)_ For each batch of specimens extracted and analysed by PCR, contamination was coded as none, some (where less than half of the blank controls were contamin ated) or heavy (where half or more blanks were contaminated). The number of samples tested and the number of positive samples were recorded for each batch, and the peR sys tem noted. The s tatis ti ca l ana lysis consisted of regarding the number of pos itive samples as the number of 's uccesses' obtained in binomial trials, each performed with a constant probability of success. This probability was then related to the two exp lanatory factors: PCR system and level of contamination. Generalised linear models (GUM; Aitken et ai., 1989), analogous to simple linea r regression for ordi­ nary quantita ti ve data, were used , w hich provid ed a tes t of s ign ifi cance for each factor (peR sys tem, level of con tamination), by giv ing a tes t sta tistic which can be assessed by comparison wi th the 0.01 critical value of the X2 distribution_ The 0_01 level of significance was set, in this case, in view of the five comparisons being made. In this way a value of grea ter th an 6.64 (d! = 1) indicated that, for the individual compa risons, it is more than 95 % probable that there is a relationship between successes and the factor being examined. It was clear that the success rates were different fo~ different peR sys tems (test s tatis tic 12.49, P < 0.Q1, allowing for multip licity of testing), with most successes being achieved with mtDNA region V PCR. Amongst the archaeological samples, only those from the United Kingdom showed any dependence on the level of contam­ ination (even making suitable allowance for multiple testing, since five data subsets - batches - were analysed). In these samples it was clear that the chance of obtaining a positive result was sign ifican tly less for the higher levels of contamination (test s ta tis tic 11.63, P < 0.01). No significant relationship with level of contamina­ tion was detected for specimens from Central and Southern Europe (test statistic 0.27, P> 0.1) or

Evison, M.P., Fieller, N.R.J. and Smillie, D.M. (1999). Ancient HLA: a preliminary survey. Ancient Biomolecules, 3(1), 1-28. ISSN 1358-6122

6

lvLP. EVISON e/ al. TABLE

v Summary of archaeological results (irrespective of contamination) HLA-OPBl

Amelogenin

Region

n

mtDNA region V

Total

+

++

%

+

++

%

+

++

%

11

+

7+

%

18 7 0 8

44

4 10 18 32

3 5 4 12

64 37 55 48

3 12 7

0 19 6 25

27 76 33 51

33 123 120 276

8 33 27 68

4 31 10 45

36 52 3) 41

Central/Southern Europe Greece Britain

41 40

11

Total

92

14

11

2

5 24

22

n = total number of specimens; + = specimens for w hich one of the du plicates gave PCR products; both of the duplicates gave PCR products; % = percentage of positive specimens

TABLE VI Region

Central/Southern Europe Greece Britain Total

Summary of archaeologica l results (no contamination in any PC Rl

HLA-OPBl

Amelogenin n

24 2

27

++ = specimens for which

+ 0 9 0 9

.l

.~.

0 4 0 4

mtDNA region V

- ---%

+

++

%

+

---r

0

1 5 0 6

0 1 0

100 25 0 26

0 4 0 4

0 13 1 14

54 0 48

Total

%

Greece (test statistic 0.21 , P> 0.1 ). For the foren­ sic blood spot samples, there w as some slight evidence that success was posi tively associated with the level of contamination (tes t value 3.86, p ~ 0.001). Data used in the GUM analysis are included in summary form in Appendix III.

DISCUSSION Modem Samples It is possible to assess sex on the basis of skeletal morphology with a maximum reliability of 95%, and normally rather less than this figure (Bass, 1987). The error level of 3% (or less) experienced in DNA sexing in this study is unsatisfactory fo r archaeological interpreta tions and any error would be critical in forensic cases. We believe the discrepancy between presented sex and amelo­ genin sexing results represents human error in

+

. .

%

1 18 0 19

0 18 1 19

33 50

0

3

71

72

50 67

6 81

+ = specimens for which one of the duplicates gave PC R products; both of the duplicates gave PCR products; % = percentage of positive specimens.

n = total number of specimens;

n

++

=

17

47

specimens for which

sampling or recording, but it was not possible to resolve apparent sexing errors by retesting fresh samples as ethical considerations preclude fur­ ther blood samples being sought from specific donors solely for research purposes. Human XY females (Jager et al:, 1990) and XX males (de la Cl;.apelle et al., 1984; GueJlaen et al., 1984) have been described, however, and chromosomal anomalies (see Schafer and Goodfellow, 1996; H aqq and Donahoe, 1998 for recent reviews) could potentially explain some erroneous results. A possibility that donations included in the sample may ha ve originated from trans-sexual or transvestite (cross-dressing) donors provides an important reminder that biological sex and gender a re not necessaril y the same, and cannot be presumed to have been the same in the past (see Brown, 1998 for a discussion). Located on chromosome 6, autosomal loci such as those of HLA would not normally be expected to exhibit sex linkage. The finding of a

Evison, M.P., Fieller, N.R.J. and Smillie, D.M. (1999). Ancient HLA: a preliminary survey. Ancient Biomolecules, 3(1), 1-28. ISSN 1358-6122

ANCIENT HLA: A PRELIMINARY SURVEY

statistically significant difference (at the 0.05 level) in the ratio of males to females in the HLA-B27 positive group is therefore unexpected The relationship_ between HLA-B27 and the ankylosing spondylitis pathology is well known (Brewerton et ai., 1973; Schlosstein et ai., 1973), with the HLA-B27 allele being present in more than 95% of primary ankylosing spondylitis patients. The association is as strong in females as it is in males (Tiwari and Terasaki, 1985: 85) Ankylosing spondylitis has long been diagnosed at a higher frequency in male patients, however, and is still believed to occur more frequently in males, at a ratio of circa 0.7"1 (Cal in, 1994) This value is close to the ra tio (0.69) of males to females in the HLA-B27 positive group in the sample tested in this study, and is well within the 95 % confidence interval for the population. This raises the pOSSibility that the ratio of male to female ankylosing spondylitis patients reflects the underlying frequency of HLA-B27 in males and females in the population as a whole. The reason for the difference in frequency of ankylosing spondylitis in males and females is not known. Shirakura et al. (1977) have suggested that an X-linked factor may be responsible A difference between males and females in HLA and ocular disease susceptibility has also been postulated (Konno et ai., 1996) and the influence of sex on the penetrance of HLA-associated rheumatoid arthritis continues to be the subject of investigation (Meyer et al., 1996). The impor­ tance of environmental factors in HLA-associated disease is well known (Tiwari and Terasah 1985) and the difference in frequency in males and females may reflect gender-based differences in social or occupational patterns. Differential diagnOSis has been thought to be partly asso­ ciated with males having a greater propensity to disclose back pain. In this study the sample was taken from a healthy, but selected group of bone marrow donors. Assuming no overall difference in HLA-B27 frequency in males and females in the region, the greater frequency of HLA-B27 positive males in the healthy sample

7

could therefore imply that HLA-B27 positive females in this region are, paradoxically, exclud­ ing themselves from the donor register on health grounds more frequently than males. An imbal­ ance in male and female donors from white and African or Afro-Caribbean communities could produce a similar effect: HLA -B27 alleles which tend to be found in African communities do not all have the clear association with anky­ losing spondylitis. It is assumed that the observed imbalance in the sample in this study is merely a statistical artefact. Nevertheless, the possibi­ lity is raised that differences in the frequencies of HLA-associated diseases in males and females could sometimes be the result of underlying gene-frequency differences in discrete local populations. The overall frequency of HLA-B27 in the sam­ ple (8.1 %) is typical for a European Caucasian population (Tiwari and Terasah 1985: 10; Calin, 1994: 146) A slightly reduced frequency of the allele might be expected in a population of healthy donors, but as Calin and Fries (1975) have established, ankylosing spondylitis can often be detected radiologically in apparently healthy HLA-B27 positive individuals. Compar­ ison of the PCR-SSP results with the serotypes of the donors revealed an exact correspondence between the DNA-based and serological results, confirming the reliability of the DNA method. Forensic Samples As expected, shorter PCR products predomi­ nated, with the relative length of the mtDNA product (compared with the amelogenin prod­ ucts) being compensated for by its higher copy­ number. Similar success was achieved with amplification of mitochondrial and amelogenin sequences. HLA-OPBl sequences were success­ fully amplified roughly half as commonly as the other two products. This trend tended to prevail irrespecti ve of the source ma terial, but in fixed bone cross-sections and teeth, HLA-OPBl and mtDNA region V sequences were relatively more

Evison, M.P., Fieller, N.R.J. and Smillie, D.M. (1999). Ancient HLA: a preliminary survey. Ancient Biomolecules, 3(1), 1-28. ISSN 1358-6122

8

M.P. EVISON et al.

prevalent than might have been predicted. Rela­ tively high HLA-DPBl recovery in bone cross­ sections may be a consequence of formalin treatment. Fixingin 10% formalin may preserve longer DNA strands, but at the same time hinder DNA purification, .especially in teeth. In four cases absence of amelogenin sequences did not preclude amplification of HLA-DPBl and in six cases amelogenin DNA was amplified, but not mtDNA In three of the latter cases an incorrect DNA sex was obtained, however. In the sample as a whole, amelogenin sexing was correct in 64 out of 84 cases (76%) where the sex of the donor of the specimen was known (Table IV). An improvemen t in results was detected when the refined method was applied in the final study of 82 specimens. Here, amelo­ genin sexing was correct in 47 out of 57 cases (82%) where the sex of the donor was known. The sex of individual donors of the bone cross­ sections was not known . Whilst the sexing results obtained (5 males and 8 females) were not inconsistent wi th the know n proportion of males to females in the sample (31 males and 27 females), some 'false female' results may be evident. Known amelogenin sexing errors were exclusively 'false females'. We have observed apparent female results from known male mate­ rial in previous exp eriments (results not shown) and attribute these to preferential amplifica tion of one target strand, normally of the shorter (X chromosome) product, in highly degraded or 'dirty' samples. Stone et al. (996) suggest that in some samples amplifiable DNA quantity may fall to a level approaching equivalence to a haploid genome. In such cases, a great deal of chance variation w ill influence whether the X or Y sequence tends to be amplified in males (in their experimental comparison it again tends to be the X): in some samples there may simply be too little DNA to confidently sex the individuaL In all but one of the 'false female' cases we encountered in this study HLA-DPBl sequences have failed to amplify and mitochondrial region V sequences are also absent or faint, confirming

the degraded nature of the samples and the low, or effectively low, concentration of DNA We have found that samples extracted by this method rarely contain sufficient inhibitors to prevent amplification of a modern DN A sample. Nevertheless, we note that 'false female' results predominated in the partially decomposed for­ ensic material or those bones and teeth w hich had been fix ed in 10% formalin. We found that one forensic bone specimen (CC1; from a two-year­ old partially skeletonised body) gave such a result, but when the DNA sample was put through a second purification step, a clear male DNA sex was obtained. Two other specimens gave the correct result when the extraction pro­ cess was completely repeated (others could not be repeated due to lack of material). In some cases, however, a second silica purification of the in itial DNA extra ct resulted in the loss of any amplifiable DNA, probably due to the 'overhead' inherent in the silica method . Stone et al. (1996) avoided the potential problem of preferential amplification of the shorter s trand by ampli­ fying a different 112 bp region of amelogenin occurring without deletion on both the X and Y chromosomes. Dot-blots were used to test for the X and Y homologues, w hich have slightly dif­ ferent sequences. Dried blood spots on gauze were an especially useful substrate and all of our specimens, w hich were up to 26 years old, yielded amplifiable nu­ clear DNA allowing amelogenin sexing, which w as confirmed to be correct from records. Not surprisingly, all recent post-operative bone speci­ mens tested were also amenable to amelogenin sexing, and HLA-DPBl and mtDNA region V an alysis. Good results were obtained from unfixed tooth fragments, but fixed teeth yielded poorer results. Although the DNA content of teeth may be rela­ tively low compared with bone, the tooth struc­ ture itself offers physical protection to DNA preserved inside. The proportion of organic mate­ rial resulting from diagenesis is less and will interfere less with DNA purifica tion. Treatment

Evison, M.P., Fieller, N.R.J. and Smillie, D.M. (1999). Ancient HLA: a preliminary survey. Ancient Biomolecules, 3(1), 1-28. ISSN 1358-6122

ANCrENT HLA: A PRELIMI.t'iARY SURVEY

with 10% formalin inhibits DNA recovery to a greater extent in teeth than in bone, perhaps reflecting a greater extent of fixing. Skeletal material included specimens which had been subjected to a variety of chemical processes as a result of the burial environment or post-mortem preserva tion procedures. Burial in quick-lime, which might be expected to accelerate DNA degradation by oxidation, does not seem to have too adversely affected DNA preservation. In fact, burial in quick-lime may enhance DNA preservation (see below). Results were also obtained from skeletal material which had been cleaned by immersion in papain solution, fixed in formalin or cleaned by prolonged immersion in hot water (95' 0 and stored at room temperature for up to 15 years. Measures of general organic preservation have been considered as indicators of DNA surviva l in archaeological ma terial (Poinar et nl, 1996) and may provide a similar indicator in forensic specimens. Jarvis (1997) measured the nitrogen content of 16 specimens of the exhumed skeletal material, up to 91 years old, included in this study. We recovered DNA from 15 of the 16 sam­ ples. It was not possible to associate the failed sample with diminished nitrogen content. The relationship between DNA preservation and nitrogen content may not be straightforward, however (Colson et ai, 1997). One specimen (Al) had been buried in a grave with quick-lime and later contaminated with engine oil. DNA was recovered, nevertheless, and Jarvis (997) pOints out that lime treatment may enhance organic preservation by inhibiting microbial and bio­ chemical activity. Age is not the determining factor of single copy nuclear DNA survival, an observation which Richards et al. (1995) have made with reference to mtDNA survival in archaeological material. Previous s tudies of forensic specimens have indicated that soiled materials tend to be poorer sources of suitable DNA than clean specimens and that putrefied biological material may be unsuitable for analysis (Gill et ai, 1985, 1987;

9

McNally et aI., 1989a,b; Balazs et aI., 1990; see also Wilson et ai., 1995a,b). The contrasts between results obtained from dried blood spots (stored on clean gauze), fixed specimens and forensic skeletal material examined in this study are con­ sistent with these observations. A high level of random contamination of extraction and PCR blanks was experienced in the study, w ith contamination eviden t in 38% of all blank controls analysed. Does this mean that the positive results can be explained by a pattern of random contamination? It is not likely that 82% of specimens sexed correctly by chance. Also, although we sometimes experience false female results with degraded forensic or archae­ ological material (and male control DNA diluted to below 25 pg [11- 1), false male results (i.e. with two bands) are extremely rare (Evison, 1996). In this study there were no 'false males', which would be predicted if random contamination of the forensic DNA sa mples was occurring. The GUM analysis showed that positive results are not at all correlated with contamination of the blanks, as would be expected if random labora­ tory-borne contamination was the explanation of the positive results. A slight association between positive results and the level of contamination of blanks was evident in the dried blood spot sample only. Given the 100% success obtained in amelogenin sexing of these specimens, we suggest that PCR carry-over is the most likely explanation for the pa ttern of contamination in this sample. Archaeological Samples The results obtained from archaeological speci­ mens must be considered in the light of extensive contamination detected in the earlier stages of the study and in the extraction, purification and silica binding blanks included in different steps in the study of the archaeological material. Although skeletal sex and the ankylosing spon­ dylitis pathology were included in the study wi th the intention of providing DNA-independent

Evison, M.P., Fieller, N.R.J. and Smillie, D.M. (1999). Ancient HLA: a preliminary survey. Ancient Biomolecules, 3(1), 1-28. ISSN 1358-6122

10

M.P. EVISON e/ al.

controls, the failure of archaeological samples to generate more than one clear band on amelo­ genin analysis in all but two cases and the failure of the two ankylC?,sing spondylitis samples (B ox Lane and Eccles) to generate nuclear DNA products meant that,no such controls could be as­ sessed The poor results ob tained with the HLA­ B27 typing are consistent with earlier results (see above) and are probably due to lack of robustness in the system when applied to ON A originating from degraded or ancient materii'lL The amelogenin results are consistent with the generation of single PCR products, usua lly cor­ responding to the shorter X chromosome target (Faerma n et ai, 1995), in archaeological and degraded forensic samples. In view of the diffi­ culties experienced by Stone et al. (1996) when using the method of Sullivan et al. (1993), it is perhaps surprising that any results were generated using this method. A comparison of PCR results obtained regard­ less of contamination (Table V) with those results obtained with a more stringent consideration of contamination, admitting only samples with no associated con taminati on in any ana lYSis (Table VI) , reveals . that 71 % of PCR results were abandoned because of contaminated blanks in one or other analysiS. Better results were obtained w ith the shorter length PCR product and with mtDNA (Table VI), which is consistent with predicted results from ancient samples (Bro wn and Brown, 1992). Even with these spec­ imens, poor ov erall consistency was evident however, with no specimen generating PCR products for all three PCR systems and both duplicate extractions (Table VII). This asks whether the positive results in uncon taminated batches are merely the result of fortuitous con­ tamina tion. Again, GUM showed that positive results are not at all correlated with contamina­ tion of the blanks, as would be expected if random labo rat ory-borne contamination, PCR carry-over or 'carrier effect' from a few archae­ ological specimens was the explanation of the positive results. It is possible that genuine

TAB.LE VII Consistency of uncontam inated positive results between PCR systems Results (11/6)

0

Central/Southern Europe

Greece

0

2 4 6

1 2 3 4 5 6

n /6

0 0 0 0 0

Britain

Total

1

3 5 7 10

10 2 0 0

0 1 0 0 0 0

2

0 0

= number of specimens out of a to tal of si x.

positi ve results are being excluded because of ass ociated contamination of negative controls. Contamination of bone would be expected to generate male amelogenin sexing results at least as frequently as female, but only two male re­ sults were ob tained from the 104 archaeological samples tested. Contamination canno t be in­ voked as a general explanation for positive results from archaeological material and the distribution of PCR product lengths in positive samples, with mtDNA and shorter genomic sequences tending to predominate, indicates that the sequences originate from the ancient material (Brown and Brown, 1992; see Table VI). Reproduc­ ibility of results (see Table VII) reflects this underlying pattern. The pOSSibility remains, however, that positive results are the consequence of in vitro recombina­ tio~ and 'jumping PCR' generating spurious PCR products. Paabo and Wilson (1988) demonstrated that PCR can generate artefacts and, as a con­ sequence of an experimental stud y of the ampli­ fication of fragmented DNA, Paabo et al. (1990) fur ther demonstrated tha t damaged DNA from an ancient sample induced partially-extended amplification products to jump from one tem­ plate molecule to another. The ability of Taq polymerase to insert adenosine a t base-less sites was also established. In order to resolve the DNA sequences ob­ tained from putative positive specimens, a small number of samples were analysed for their

Evison, M.P., Fieller, N.R.J. and Smillie, D.M. (1999). Ancient HLA: a preliminary survey. Ancient Biomolecules, 3(1), 1-28. ISSN 1358-6122

ANCIENT HLA: A PRELIMINARY SURVEY

HLA-DR, DQ type and HLA-DRBl subtype. Am­ plification refractory mutation system polymer­ ase chain reaction (peR-ARMS; Newton et ai., 1989; Tsuji et ai., 1.992) for HLA-DR, oQ is a form of peR-SSP in which a panel of 24 peR reactions are used to differentiate between haplotypes. Whilst individual reactions in the panel may prove positive (or negative) for more than one haplotype, the pattern of positive and negative results in the panel as a whole is uruque for each haplotype motif. peR-ARMS generated the same HLA type, DR 3 11, oQ 2 7, for separate samples originating from separate specimens of skeleton THE 1 (dated to circa 14,000 Be). A Neolithic skeleton (TH E 27-0,75 58) from the same excava­ tion (Theopetra Cave) gave the same HLA-oR, oQ type, however, although the result was extremely faint (see Fig. 9). The experimenter's HLA type is DR 2 7, oQ 2 6. The recovery of the same HLA-DR, DQ type from separate samples from the same Upper Palaeolithic individual is

11

encouraging, although the detection of the same HLA type in a Neolithic individual from the same excavation raises the possibility of back­ ground contamination with DR 3 11, oQ 2 7 DNA. Although it cannot be ruled out, back­ ground contamination is very rarely experienced as a problem in routine HLA typing from modern material due to th e vast excess of bona fide DNA from the sample being tested. It is also possible that the faint result in the Neolithic sample is caused by peR carry -over from the Upper Palaeolithic specimens or, indeed, that this sam­ ple has also typed correctly. Mixed results were obtained from the HLA­ oRBl subtype dot-blot analysis (Olerup and Zetterquist 1991; results not shown), with faint hybridisation only occurring, in some, but not all of the archaeological samples (despite the generation of a clear peR product on amplifica­ tion for generic HLA-oRBl) In addition, single samples gave faint results for more than two

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 192021 222324 M

A

B

c D

FIGURE 9 Photograph of 1.5% agarose gel showing results of PCR-ARMS typing for HLA-OR, OQ for four archaeological specimens. Rows: A-C, test specimens THE 1, THE 1/ 2, THE 27-0,75 58 respectively; D, experimenter. The OR, OQ type is indicated by the pattern of presence and absence of products resulting from amplification of 24 different primer sets (lanes 1- 24). In rows A, 8 and C the following products amp lify (indicated by arrows): .lane 4, ORBl"0301, lane 5, ORBl'0301; lane 7, ORBr1101-13; lane 17, OQBl'0201; lane 18, OQBl'0301; lane 20, OQBl'030l: corresponding to serotype OR 3 It OQ 2 7. Rows 8 and C are increasingly faint. [n row D the following products amplify: lane 2, ORBr1501; Jane 12, ORBl"0701; lane 17, OQBl"0201; lane 23, OQBl"0601-3: corresponding to serotype OR 2 7, OQ 2 6. Lane M contains molecular weight markers

Evison, M.P., Fieller, N.R.J. and Smillie, D.M. (1999). Ancient HLA: a preliminary survey. Ancient Biomolecules, 3(1), 1-28. ISSN 1358-6122

12

M.P. EVISON el al.

HLA-DRBI subtypes, with a restricted number of subtype sequences recurring in all samples. The experimenter's subtype was consistent with the PCR-ARMS r~sult. Multiplex PCR analyses for VWA31, FES /FPS, THOI and F13Al targets by D. Love and K. ~orringe (personal commu­ nication) have so far yielded inconsistent resu lts from samples extracted in this study, with a small subset of archaeological Selin pIes giving some product. A study of mtDNA in the same set of archaeological samples by I. Colson (persona l communication) detected only the experimen­ ter 's sequence and the European reference. The HLA-DRBI results are consistent wi th an in vitro recombination or jumping PCR explana­ tion for apparent positive bands, consisting of a mixture of DNA sequences, partly meaningless chimeric molecules and partly a mixture of genu­ ine sequences of uncertain origin. The HLA-DR, OQ PCR-ARMS typing products cannot be ex­ plained as recombinant artefacts as the method uses multiple specific primer sets to generate a single DR, DQ type. In vitro recombination and jumping PCR might be expected to generate results with a random selection of primer sets, leading to a meaningless result. A cautious explanation would be that the results from THE 1 are due to background contamination with DNA of a single DR, DQ type wh ich means pleading a 'special case' for contamination of these samples. Ancient HLA and Other Nuclear DNA Analyses The earliest ancient DNA publications describe the recovery of human nuclear sequences, albeit from exceptionally well preserved material such as Egyptian mummies (Paabo, 1985a,b, 1989) or the Windover brains (Doran et al., 1986; Lawlor et al., 1991). Possibilities of genuine > 10,000 bp DNA sequences occurring in ancient tissues now seem remote, however (Del Pozzo and Guardiola, 1989). Although they cannot be used as a model for single copy genes, multicopy targets offer a

greater likelihood of successful amplification. Thuesen and Engberg (1990) reported the recov­ ery of 284 bp Alu and 340 bp ex-repeat sequences from mummified tissue and bone from a 1500­ year old Inuit site in Greenland. Although negati ve controls were included in the DNA­ DNA hybridisation analysis, extraction controls were not described. Gaensslen et af. (1993) amplified 123 bp products of a homeobox se­ quence from two out of three commingled skeletal elements of 19th century origin. PCR controls, but not negative extraction controls, were described. Y-repeat methods have been applied to ordinary skeletal material, not pre­ served in peat or by mummification (Hummel and Herrmann, 1991; Hummel et al., 1992; Lassen et nl, 1994; Richards et aI., 1995), but suffer from the lack of a positive test for the X chromosome. Single copy nuclear sequences have been targeted Meijer et al. (1992) described the ampli­ fication of sequences of the X-chromosome factor VIII intragenic fragment (142 bp) and of exon 7 of the phenylalanine hydroxylase (PAH) gene, of 155 and 107 bp respectively, generated using a nested PCR. Filon et al . (1995) amplified a thalassaemic variant of the ,G-globin gene in samples from a skeleton exhibiting classic patho­ logical indicators of chronic anaemia. Only one of six archaeological samples yielded a positive result, prod ucing a' thalassaemic variant of the ,6'-globin gene sequence. No control of skeletal material free of 'anaemic' symptoms was in­ cluded. Quantities of bone tested were small for ancient DNA work, at . . . , 0.5 and . . . , 1.0 mg. Parr et al. (1996) reported the recovery of a 387 bp sequence of exon IVof the genomic human lectin gene from two skeletal specimens. No hy­ bridisation or sequencing analysis of these prod­ ucts was undertaken. Beraud-Colomb et al. (1995) analysed archaeological bone specimens for sequences of the ,6'-globin gene and obtained amplifica bon prod ucts for 200-515 bp targets in five of ten samples tested. Reproducibility was limited and a 678 bp target failed to amplify. Faerman et al. (1995) carried out a s tudy of DNA

Evison, M.P., Fieller, N.R.J. and Smillie, D.M. (1999). Ancient HLA: a preliminary survey. Ancient Biomolecules, 3(1), 1-28. ISSN 1358-6122

ANCIENT HLA: A PRELIMINARY SURVEY

sexing using amelogenin gene targets of 218 and 330 bp. Six out of seven specimens sexed correctly by comparison with the skeletal sex. Sex was assigned to a fu[ther eight out of twelve speci­ mens of unknown sex. The presence of only the longer X chromoso.me product was assumed to be a genuine female result, not a 'false female', because of preferential amplification of the shorter Y product. X and Y chromosome specific DNA sequences were established for only one sample. Stone et al. (1996) correctly de termined the sex of 19 out of 20 skeletons from the Norris Farm Amerindian site, assessed by comparison with the skeletally-determined sex. On average, 2.4 amplifications are necessary to obtain any product from the samples . Very often four amplifications were needed to establish within reason that no Y chromosome target was present. Products from only a single chromosome were detected in 24 out of 39 amplifications. In 14 out of these 24 cases it was the X chromosome only which was detected. Attempts to amplify DNA from this material using the method of Sullivan et al. (1993) were not successful. Ramos et al. (1995) examined dinucleotide repeats in 28 sam­ ples of bones and teeth of up to 5000 years of age. Discrepancies were found in 97% of samples tested. The same genotype reoccurred in 46% of the samples. Whilst extraction or PCR controls are not discussed, the lack of consistency in the results obtained tends to mitigate against con­ tamination with intact modern sequences. The amplification of HLA targets has also been reported. Lawlor et al. (1991) amplified 68-107bp segments from a sin gle Wind over specimen and were prepared to tentatively assign typings of HLA-A19 and HLA-B37. Further research on the Windover material (Hauswirth et al., 1994) yielded 123-135 bp single copy APO-A2 micro­ satellite sequences in six samples tested. Targets of 68-107 bp from the HLA class I third exon amplified in 14 specimens (at least 15 were tested) and all hybridised with at least one of the spe­ cially designed allele-specific oligonucleotide probes. Interestingly, the HLA-A19 allele was

13

present in all 14 samples. Ivinson et al. (1993) were able to assign HLA-ORBl types to three ap­ parently related mummies from the Manchester Museum collection by analysing a short, 71 bp, section of the HLA-ORBl second exon, but stated that "contamination and false positive results must not be underestimated". Woodward et al. (1994) compared teeth and soft tissue samples from 20 different mummies. Products of 239­ 242 bp, from the polymorphic HLA-OQ Al locus, were obtained from 18 out of 20 teeth. Elec­ trophoretic single-strand conforma tion al poly­ morphism (SSCP) analysis was performed on the 18 samples and 12 HLA-OQAl alleles were distinguished. HLA-OQAl sequences were ob­ tained from 18 subclones from three haplotypi­ cally-differen t individuals distinguished by SSCP. The results generated two allelic variants, types 1.3 and 4. Of the three individuals, one was a 1.3 / 4 heterozygote and the other two were homozygous for type 4. This cursory survey of the literature reveals that single copy nuclear DNA products have been generated from 74 out of 92 archaeologi­ cal teeth and bone samples tested: Meijer et al. (1992), 15 out of 20; Woodward et al. (1994), 18 out of 20 teeth; Filon et al. (1995), one; Beraud­ Colomb et al. (1995), five out of ten; Parr et al. (1996), tw o out of two; Stone et al. (1996), 19 out of 20; Faerman et al. (1995), 14 out of 19. Ramos et 91. (1995) obtained sequences from seven out of 28 bones and teeth, but regarded their results as demonstrating "more inconsistency than use­ fulness " . Given that sample sizes have been too small and results too equivocal for statistical analysis or phylogenetic study, the success rate is likely to appear substantially higher than it really is. Only 25 of the 74 'positives' have been corroborated by DNA-DNA hybridisation analysis (19 samples) or sequencing (6 samples ). A number of papers omit essential controls. Although contamination is generally recog­ nised as a problem, the ex tent and pattern of contamination and excluded results are rarely reported.

Evison, M.P., Fieller, N.R.J. and Smillie, D.M. (1999). Ancient HLA: a preliminary survey. Ancient Biomolecules, 3(1), 1-28. ISSN 1358-6122

14

M.P. EV[SON et al.

In a comparable survey of short tandem repeat (STR) su rvival in 200 Medieval skeletons, Zierdt et al. (1996) described the successful amplification of.DNA ob tained from 36.0% of tooth and 22.2% of bone specimens examined. Negative control sqmples and internal controls for assessment of reproducibility were included, but neither the extent of contamina tion nor the deg ree of reproducibility experienced were reported. The authors acknowledge that false­ positi ve results due to contamination cannot be ruled out in indi vidual cases, but argue that the extent of heterozygosity at the VWA31/A locus detected in 76 of the 200 specimens was consistent with the bona fide recovery of ancient DNA in the sample as a who le. Prob­ lems with size classification of alleles on the basis of electrophoretic mobility and the genera­ tion of PCR artefacts (shadow bands), even from mo dern substrates, are recognised and, given these difficulties, the influence of in vitro recombination or jumping PCR w ith ancient targets is likely to be significant, but is not considered. Sequencing of some samples from the investigation was reported, but the authors indicate only th at the results agreed w ith the sizing assessed by polyacrylamide gel electrophoresis. Returning to the results obtained in this s tudy, difficulties in applying amerogenin sexing to arch aeological material (Faerman et al., 1995; Stone et aI., 1996) may explain why clear DNA sexing results could not be obtained using the amelogenin method of Sullivan et al. (1993), even though some genuine sequences might be present. At 327bp, the HLA-DPBl target may be at the upper limit for ancient nuclear DNA survival, although successes with longer sequences have been reported. Thuesen and Engberg (1990) described the recovery of 340 bp multicopy nuclear sequences, and Beraud­ Colomb et al. (1995) and Parr et al. (1996) reported the recovery of single copy genomic sequences of 377,51 5 and 387 bp, respectively Faerman et al. (1995) targe ted an amelogenin X chromosome

homologue of 330 bp . The presence of up to four HLA-DPBl subtypes in extracts from single speci­ mens is consistent with the generation of artefacts in the early stages of PCR (Paabo and Wilson, 1988; Paabo et al., 1990; Ramos et al., 1995) The presence of a restricted number of subtypes may indica te that certain recombinant molecules are generated much more efficiently than others. Interestingly, Hauswirth et al. (1994) found a particulClf HLA variant (HLA-A1 9) to be present in all 14 of their samples from Windover an d Woodward et al. (1994) found HLA-DQAl type 4 subvariant alleles to be presen t in all three individuals they were able to type. The HLA sys tem contains a multiplicity of related poly­ morphisms at numerous loci, including several pseudogenes (Charro n, 1997) . Gene duplication and gene conversion events occurring during the evolution of HLA could account for the generation of 'meaningful' recombinant mole­ cules during amplification from an ancient DNA substrate.

Further Work

We hope to undertake further study of mate­ rial frow archaeo logical sites in Thessaly and Macedonia, or other sites exhibiting similar climatic and geomorphological conditions, incor­ porating cloning and sequencing of multiple PCR products to establish the underlying pattern of ~ecombination occurring during amplification If sufficient sequences are obtained, it may be possible to deduce the genuine allelic variant sequences present in the samples to a given level of statistical probability, as Goloubinoff et al. (1993) and Allaby et al. (1999) have demonstrated for alleles of maize and whea t, respecti vely The extent of any PCR mis-priming should be assessed (Don et aI., 1991; Chou et al., 1992; Carmody and Vary, 1994). A number of our specimens have been passed on to . another HLA laboratory to establish whether the results can be independently reproduced.

Evison, M.P., Fieller, N.R.J. and Smillie, D.M. (1999). Ancient HLA: a preliminary survey. Ancient Biomolecules, 3(1), 1-28. ISSN 1358-6122

Al'\fCIENT'HLA: A PRELIMINARY SURVEY

CONCLUSIONS Although problems of contamination are well recognised in ar:cient DNA work, the extent and pattern of contamination is rarely reported. In this study, the analysis of a large number of samples with a sufficient proportion of controls present, permitted a statistical study of the pattern of contamination and allowed contam­ ination to be excluded as a general explanation for positive results obtained. We suggest that analyses of small numbers of specimens some­ times reported in the literature, often with limited controls, are not sufficient to exclude a fortuitous pattern of contamination from PCR carry-over or other sources, even though meaningful DNA sequences may subsequently be obtained . Statis­ tical verification must be supported by an indica­ tion of the validity of sequences obtained, how­ ever. Southern blotting or similar hybridisation analysis provides some indication, but cloning and sequencing of multiple PCR products is necessary to properly resolve putative positive results. The reproduction of results in an inde­ pendent laboratory is an ultimate requirement. Continual problems with contamination indicate that dedicated clean facilities, including a clean room or equivalent eql'jpment, are an absolute prerequisite for ancient DNA research. The recovery of single copy nuclear human DNA from ordinary archaeological teeth and bone is perhaps the most important goal for ancient DNA research (a small number of successes with uniquely preserved tissues are not likely to provide a useful research tool for archaeology). Results from an exceptional range of archaeological periods from excavations in Greece are promising, and hold the possibility of the study of gender, immunity and kinship in ancient populations. It would be premature to discount single copy nuclear sequences as viable ancient DNA targets. The material from Greece, in contrast to poor material from the UK, and similar evidence from the literature, leads the authors to believe that

15

relatively inexpensive subsampling and analysis of skeletal material from a wide variety of depositional environments will provide a bet­ ter (and more cost-effective) empirical gUlde to ancient DNA survival than theoretical models based on organic geochemical analysis.

MATERIALS AND METHODS Materials Modem Samples

Whole peripheral blood specimens (n = 431) taken from healthy blood donors were analysed by amelogenin and HLA-B2 7 PCR-SSP Standard salt-precipitation methods (Miller et ai , 1988) were used for the extraction and purification of DNA from large (5 m!) and small (1 m!) quantities of whole peripheral blood. The quantity of DNA in the samples obtained from these specimens was measured by spectrophotometry, in order to establish that a satisfactory yield and 'purity' (a DNA: protein ratio, assessed spectrophotomet­ rically, of 1.6-1.8: 1) was being obtained, and to allow dilution of the samples to the required concentration (25ng~I-1) for PCR analysis The HLA-B27 phenotype of the donors had been established serologically prior to this study and the presen ted sex (or gender) of the donors was also recorded. The PCR analysis was carried out without prior knowledge of these results. An initial trial of HLA-DPBl and mtDNA region V PCRs was not attempted as the suitability of these methods for popula tion (Tsuji et aI., 1992; Charron, 1997) or ancient DNA studies (Hagelberg et al., 1989, 1991; Paabo 1989; Hanni et aI., 1990; Hagelberg and Clegg, 1991) respec­ tively, is well established. Forensic Samples

The forensic samples consisted of 119 specimens ranging in age from a few days to circa 90 years (see Appendix D. The specimens included dried

Evison, M.P., Fieller, N.R.J. and Smillie, D.M. (1999). Ancient HLA: a preliminary survey. Ancient Biomolecules, 3(1), 1-28. ISSN 1358-6122

16

M.P EVISON et af.

blood spots (n = 28), extracted teeth and shed deciduous teeth (n = 18), post-operative bone fragments (n = 6), bone cross-sections preserved in 10% formalinjn = 20), partially skeletonised material (n = 10) and exhumed skeletal remains from burials (n = 37). The 119 specimens included 37 which were used as model substrates during methodological development (see below) and 82 which were used in a final study employing the refined method. Post-operative fragments of femur and tibia were recovered following routine orthopaedic surgery Exhumed material came from three cemeteries. Burials had taken place between 1904 and 1984. Specimen Al originated from a grave treated with quick-lime (CaO) and subse­ quently contaminated with automotive lubricat­ ing oil. Cross-sections of tibia had been removed from donated medical cadavers during 1980 and fixed in 10% formalin. Forensic material origi­ na ted from cases arising between 1986 and 1994. Not all material was skeletonised when recov­ ered. Forensic specimens SF and SM had been cleaned by immersion in hot water (95°C) for several days. Specimens Y, Yl and Y2 had been fixed in 10% formalin prior to cleaning in papain. Extracted permanent teeth had been collected following dental treatment and stored at 4°C. Specimens ET4 to ET8 had been fixed in 10% formalin. Specimen ES was a shed deciduous tooth stored at room temperature for 2 years. Dried blood spots had been collected on sterile ga uze over a 26 year period from 1969 and stored in sealed containers at room temperature.

originated from Late Medieval/Early Post Medi­ eval period excavations at Abingdon (1300-1540 AD); Medieval monastic sites at Bolsover and Exeter; Anglo-Saxon sites at Berensfield, Box Lane (Pontefract) and Eccles; and Neolithic sites at Kitley Caves and Torquay Material from central and southern Europe originated from Bronze Age Mokrin (former Yugoslavia) and the Neolithic site of Eszterhorvaty in Hungary Two palaeontological samples from Germany were etlso included in the study: an Eocene plessiada­ pid specimen (Halle-Wittemberg Museum) and Pleistocene material from a straight-tusked ele­ phant (Neumark-Nord). Material from Greece originated from the sites of Makrigialos, Korinos and Agios Mamas in Macedonia, and included material dating from the Neolithic to the Classical periods. The oldest human material examined dated from circa 14,000 BC originating from Theopetra Cave in Thessaly, and including skeletons dated to the Palaeolithic and Mesolithic periods, as well as disarticulated and fragmen­ tary material of Neolithic date. All of the material exhibited good gross morphological preservation assessed macro­ scopically Of the bone material, femoral shafts were generally chosen for analYSis. A small number of samples from Germany and Italy were provided in powdered form. Of the 104 archae­ ological specimens; 92 were extracted and ana­ lys.ed in duplicate. Ten specimens were recovered in situ from the archaeological excavations at Ma'krigialos. These specimens were transferred directly upon exposure into sealed containers, without being handled or washed.

Archaeological Samples The archaeological samples (see Appendix II) consisted of 104 skeletal specimens, comprising of teeth, and bone from a number of skeletal elements, donated by archaeologists from exca­ vations in England (n=41), Central Europe (n = 20) and Greece (n = 43). The material ranged in age from Palaeolithic (14,000 BC) to Late Medieval (circa 1540 AD). British material

Ethical Considerations Consent was obtained for material donated by blood donors and orthopaedic patients, and from relatives of donors of forensic material where relevant. The anonymity of donors was main­ tained throughout. Ethical approval was given by the Ethical Committee of the Northern General Hospital, Sheffield.

Evison, M.P., Fieller, N.R.J. and Smillie, D.M. (1999). Ancient HLA: a preliminary survey. Ancient Biomolecules, 3(1), 1-28. ISSN 1358-6122

ANCIENT HLA: A PRELIMINARY SURVEY

Methods

Precautions Against Contamination Precautions wer,e taken against contamination with intrusive DNA (Brown and Brown, 1992). Bone ex traction equipment was cleaned by soaking in 5.0% sodium hypochlorite solution for 1 h. Pre- and post-PCR activities were con­ ducted in separate rooms and a laminar flow cabinet was used during extraction and puri­ fica tion steps. Surfaces of bone or teeth were cleaned by washing with 0.05% sodium hypo­ chlorite solution or by abrasion with a grinding tip or drill bit, which was then discarded. Separate aliquoted stocks of reagents were used for each batch of extractions, ae rosol resistant pipette tips or positive displacement pipettes were used, and pre- and post-PCR activities were carried out in separate laboratories. Foren­ sic and archaeological samples were extracted and purified using the silica method in batches of ten or more specimens. Blank controls were included in the extraction, purification and am­ plification steps to allow detection of contamina­ tion with extraneous DNA and to provide an indication of its or-igin. At least one extraction blank (treated identically except that 05 ml sterile wa ter was used in place of an archae­ ological or forensic specimen) was included in each batch. Purification (treated identically ex­ cept that 0.5 ml sterile water was used in place of an archaeological or forensic extract super­ natant) and PCR (5 ~Ll sterile water in place of the DNA sample) blanks were additionally included in some batches. Positive controls were used in development of the methods: extraction (50111 modern blood in 450111 sterile filtered distilled water), purification (0.5 ml DNA solution at 10 ng j..Ll- 1) and amplification (5 ill DNA solution at 10 ng ill - I) controls were used. Positive con­ trols were abandoned in the analysis of archae­ ological specimens, due to problems experienced with contamination suspected to be due to PCR carry-over.

17

DNA Extraction and Purification For ON A extraction and purification we used an adaptation of the silica method (Boom et al., 1990; Hbss and Paabo, 1993), which we refined using material of the kind described in this study as an experimental model (see above). In refining the method , w e have endea voured to retain its simplicity and rapidity, and not to increase the number of steps We have made the following observations (Evison, 1996). Use of a detergent (Triton X-lOO or sodium dodecyl sulphate) makes no difference to the effectiveness of the method, but addition of proteinase 1< is marginally advantageous. Use of 0.5 M Na2EDTA increases DNA yield, but also results in the extract contain­ ina a plethora of proteins and other biomolecules. DNA yield improves with up to 48 h of mixing in extraction buffer at room temperature, after which further gains are marginal. Ex traction temperature makes little difference: increased release of DNA may be countered by increased DNA degradation and reduced proteinase activ­ ity. We found 4 M guanidine isothiocyanate (GuSCN) superior to 6 M sodium iodide as the DNAIsilica binding agent and a 2 h binding time at room temperature to be optimal; others have found a higher binding temperature more effec­ tive, however (Smith et al., 1995). The GuSCN solution should be freshly prepared and stored in a light-proof container. A thorough washing step, prior to elution of DNA from the silica, is also important. We use a reduced silica volume (20 111), which we consider adequate for the DNA quantities involved, and elute once into a large aliquot (115111) of sterile filtered distilled water, taking off a slightly smaller aliquot (l05 il1) to avoid inadvertent removal of silica. Presence of 160 J..lg ml- 1 bovine serum albumin (BSA) in the PCR reaction made no difference to product yield, perhaps reflecting the effectiveness of the silica method in removing inhibitors. A degree of random contamination of ex traction, purification and amplification blanks was experienced. When the amplifications were repeated using a layer of

Evison, M.P., Fieller, N.R.J. and Smillie, D.M. (1999). Ancient HLA: a preliminary survey. Ancient Biomolecules, 3(1), 1-28. ISSN 1358-6122

18

ME EVISON el al.

25 ~l mineral oil to prevent evaporation of the reaction mixture this problem diminished. Pre­ liminary experiments with extended PCRs (of up to 50 cycles) increased the frequency of positive results obtained, but an increased frequency of contamination was, also experienced. Extended­ cycle PCRs may be necessary if DNA is to be detected in difficult forensic or archaeological specimens, however. PCRs of 35 or 40 cycles were therefore employed and blank controls were included in all recovery and amplification steps. Attempts to apply the HLA-B2 7 PCR system to forensic specimens were unsuccessful. HLA -B27 analysis of archaeological specimens was there­ fore restricted to two skeletons exhibiting anky­ losing spondylitis (Box Lane and Eccles). Although our method is effective in releasing DNA from bone specimens and removing in­ hi.bitors, there is an 'overhead' of DNA loss which may be greater than that involved when the phenol-chloroform method (Sam brook et ai., 1989) is used. The beneficial effects of proteinase K, a high Na2EDTA molarity and lower extraction temperature may be a reflection of the types of substrates we have examined, which we expect contain a relatively high proportion of fixed pro­ teins and other biomolecules, and their break­ down products. These preliminary results were used as guidelines for the amendment of our ex­ traction and PCR protocols. Specimens of shaved post-operative bone were pulverised prior to extraction, except for one specimen, which was extracted from a femoral head by drilling with a flamed bit. Powdered bone from 16 of the exhumed skeletons was generated during the removal of segments from the shafts of long-bones by hacksaw for nitrogen analysis (Jarvis, 1997) Bone powder was ex­ tracted from the fixed bone cross-sections by drilling with a flamed bit, material from the outer surface being discarded. Other bone surfaces were cleaned by abrasion or by washing in 0.05% sodium hypochlorite solution and the bone ground to a fine powder in a coffee mill (Philips HR2811) or in liquid nitrogen using a pestle and

mortar. Intact whole teeth were prepared by washing in 0.05% sodium hypochlorite solu­ tion and fractured longitudinally to expose the pulp cavity. Dried blood spots were cut into 0.5 x 1.0 cm strips. Quantities of "'-'1.0g of bone, ",-,0.1-0.7g of tooth fragments or ,..." 0.5 cm 2 of dried blood spot (on gauze) were combined with 2.0 ml 0.5 M Na2EDTA pH 8.0 and 25 ~tl proteinase K (20 mgml-l) in 3.5ml polystyrene tubes (Sa rstedt) and mixed on a rotary mixer for 48 h at room temperature. Substrate residues were pelle ted by centrifugation at 4000 x g for 5-15 min. Aliquots of 0.5 ml extract supernatant were bound to 20 ,Ltl si lica (Sigma) suspension using 1.0 ml4 M GuSCN (Sigma) by mixing for 2 h in 1.5 ml polypropylene tubes on a rotary mixer at room temperature. DNA / silica matrix was pelle ted by microcentri­ fugation for 20 s at 13,000 x g and washed twi.ce in 1.5 ml 70% ethanol and once in 1.5 ml acetone by vortexing briefly to resuspend the matrix. The pellet was dried at 56 DC for 5 min in a heating block. DNA was eluted from the silica into 115 ,ul sterile filtered distilled water by heating at 56DC for 15 min in a heating block, and vortexing every 5 min, to aid so lution of the DNA Silica was pelle ted by centrifugation at 13,000 x g for 2 min. Volumes of 105 ~l DNA solution were taken off to avoid inadvertent removal of unwanted silica. Specimens were stored in 0.5 ml polypropylene tubes at _20DC prior to PCR analysis. peR Analysis

All PCRs are of the sequence specific primer (SSP) type and were carried out using a Perkin-Elmer GeneAmp 9600 thermal cycler. Primer sequences are given in Table VIII. PCR reaction mi.xtures and programs were optimised using control samples of modern DNA extracted from whole peripheral blood specimens. The amelogenin PCR targets the X-Y homologous amelogenin gene. The X homologue contains a 6 bp dele­ tion, generating 106 and 112 bp products from the X and Y chromosomes, respectively (Sullivan

Evison, M.P., Fieller, N.R.J. and Smillie, D.M. (1999). Ancient HLA: a preliminary survey. Ancient Biomolecules, 3(1), 1-28. ISSN 1358-6122

ANCIENT HLA: A PRELIMINARY SURVEY TABLE VIII Primer Amel-A Amel-B OPB-AmpA DPB-AmpB A

B E915 E1 36 ~5

C3

C5

Oligonucleotide primer sequences Sequence (5'

~3')

CCC TGG GCT CTG TAA AGA ATA GTG ATCAGA GCT TAA ACT GGG AAG CTG GAGAGTGGCGCCTCCGCTCM GCC GGC CCA AAG CCC TCA CTC ATG CTA AGT TAG CIT TAC AG ACA GIT TCA TGC CCA TCG TC GGG TCT CAC ACC CTC CAG AAT CGG CGG TCC AGG ACG T TGCCAAGTGGAGCACCCAA GCA TCT TGC TCT GTG CAG AT

et ai, 1993) Aliquots of 5 pi of DNA sample were amplified in a 10 !..d reaction mixture consisting of 10 mM Tris-HCl pH 8.3, 50 mM KCI, 1.5 mM MgCl 2 6H 2 0, 1 % Triton X-l00, 200 ~M dNTPs, 1 U Taq polymerase (Advance Biotechnologies) and 1.0 ~lM each oligonucleotide primer (Amel-A and Amel-B; BH Bio). The PCR program con­ sisted of an initial strand separation step (94°C, 120 s), followed by a biphasic amplification with a 10 cycle stringent phase (94°C, lOs; 65°C, 60s) and a 30 cycle non-stringent phase (94°C, 10 s; 62°C, 40 s; n°c' 30 s). The PCR products were visualised by electrophoresis . Aliquots of 4 III DNA solution were added to 1 III X5 gel loading buffer (40% sucrose, 0.1 % bromophenol blue, 50mM Na2EDTA, 'SOmM Tris-HCI pH 7.6,5% SDS) and run in a 5% agarose gel (125 V, 40 min) containing 10 III ethidium bromide 00 Ilg me l). The HLA-DPBl primers amplify a 327 bp se­ quence of the polymorphic second exon (Kimura and Sasazuki, 1992). The mtDNA analysis, in­ cluded to provide a comparison of nuclear and mtDNA survival, used primers which target a 121 bp segment of the non-coding region V (Wrischnik et ai , 1987). In both cases, aliquots of 5)11 of DNA sample were amplified in a 10 ~I reaction mixture consisting of 10 mM Tris-HCl pH 8.3, 50 mM KCl, 1.5 m.tv! MgCh.6H 2 0 , 100)..lg ml- 1 gelatin, 200)..lM dNTPs, 1 U Taq polymerase and O.5)..1M each oligonucleotide primer (DPB-AmpA and DPB-AmpB for HLA­ DPBl; A and B for mtDNA region V). The PCR program consisted of an initial strand separation

19

step (95°C,S min), a 35 cycle amplification phase (95°C, 1 min; 55°C, I min; 72°C, 1 min) and a final elongation step (72°C,S min). The PCR products were visualised by electrophoresis (100 V, 20 min) on a 1.5% agarose gel. The HLA-B2 7 primers amplify a 135bp se­ quence of ex on 3 of the polymorphic HLA-B27 allele (Dominguez et aI. , 1992) Aliquots of 3 ~I of DNA sample were amplified in a 10 ~d reaction mixture consisting of 10 mM Tris-HCI pH 8.3, 50mM KCl, 1.5mM MgCl 2 6I-hO, 100pgml- 1 gelatin, 200)..lM dNTPs, 1 U Taq polymerase and 0.5)..1M each oligonucleotide primer (E91s and EI36as). A 796 bp conserved sequence of HLA (primers C3 and C5, each at 0.24 )..1M) was ampli­ fied as a positive control for the amplification of human DNA in HLA-B27 negative samples (Olerup and Zetterquist, 1991). The PCR program consisted of a 30 cycle amplification phase (94°C, 20 s; 63°C, 50 s; noc, 20 s). The PCR products were visualised by electrophoresis (100 V, 20 min) on a 1.5 % agarose gel. Dilution series of modern DNA templates were used to assess the sensitivity of the PCRs (results not shown). The lowest concentration of modern template reliably amplified was 2.5 pg )..11- 1 for HLA-DPB1, 250 fg 111- 1 for mtDNA region V, 250 pg )..11- 1 for the HLA C3 + C5 conserved con­ trols and 2.5 pg 111- 1 for HLA-B27. Sensitivity of the amelogenin system varied for male and fe­ male substrates (the lowest concentration of mod­ er~ template reliably amplified was 25 pg )11- 1 for male DN A and 250 fg )..11- 1 for female). Only HLA­ B27 PCR-SSP typing was in routine use in the laboratory during the period of this study. Statistical Analysis A statistical analysis of the results from the modern sample was undertaken to establish the significance of the frequency of HLA-B27 ob­ served in this study, using X2 and confidence interval calculations (Gardner and Altman, 1986). Statistical significance was set a priori at the 0.05 level.

Evison, M.P., Fieller, N.R.J. and Smillie, D.M. (1999). Ancient HLA: a preliminary survey. Ancient Biomolecules, 3(1), 1-28. ISSN 1358-6122

20

M.P. EVISON et al.

Generalised linear interactive modelling analy­ sis (GUM; Aitken et aI., 1989) was used to assess the significance of random contamination in forensic and arcl}aeological samples. In order to take into account the multiple comparisons being made using GLIrvt the statistical significance was set a priori at a nominal 0.01 level to achieve an overall level of < 5%. Acknowledgements

Samples have been generously provided by Manthos Bessios and Maria Pappa (Thessaloniki Museum), Prof. A. Farneti (Istituto di Medicina Legale e delle Assicurazioni, Milan), Andrew Chamberlain and David Jarvis (University of Sheffield), Beth Rega (Claremont Colleges, Los Angeles), Jorg Erfurt (Halle-Wittemburg Museum), Birger Herzog (Friedrich-Schiller­ Universitat Jena), Nina Kyparissi (Ephoria of Speliology and Palaeoanthropology, Athens), Dietrich and Ursula Mania (Forschungstelle Bilzingsleben, Jena), Charlotte Roberts (Univer­ sity of Bradford), Bob Sokol (Trent Centre, National Blood Service), Tom Smith (Northern General Hospital), Harry Stern (Sheffield University Health Service) and Jenny Wakely (University of Leicester). This research was supported by the UK Natural Environment Research Council, the University of Sheffield and Trent Centre, National Blood Service, where the laboratory work was carried out. References Aitken, M., Anderson, D., Francis, B. and Hinde, J. (1989) Statisllcal Modelling in GUM. Oxford University Press, Oxford. Allaby, R.G., Banerjee, M. and Brown, TA (1999) Evolution of the high-molecular-weight glutenin loci of the A, B, 0 and G genomes of wheat. Genome 42, 296-307. Balazs, I., Baird, M., McElfresh, K. and Shaler, R. (] 990) Experimental techniques for the isolation and analysis of DNA from forensic materials. In, DNA in Forensic Science: Theon), Techniques and Applications G. Robertson, AM. Ross and L.A Burgoyne, Eds.), pp. 60-73, Ellis Horwood: New York. Bass, WM. (1987) Human Osteology: a Labomtory and Field Manual. Missouri Archaeological Society, Columbia. Beraud-Colomb, E., Roubin, R., Martin, J., Maroc, N., Gardeisen, A, Trabuchet, G and Goossens, M. (] 995)

Human ,6-globin gene polymorph isms characterised in DNA extracted from ancient bones 12,000 years old. Am. f. Hum. Genet. 57, 1267-1274. Boom, R, Sol, C.J.A, Salimans, MMlvl, Jansen, c.L., Wertheim-van Dillen, P.M.E. ilnd Viln der Noordila, L. (1990) Rapid and simple method for purification 01 nucleic acids. f. Clin. Microbiol. 28, 495-503. Brewerton, D.A, Caffrey, Iv!., Hart, FD, Jilmes, D.C.O., Nicholls, A and Sturrock, R.D. (1973) Ankylosing spondy­ litis and HLA-B27. Lancet i, 904-907. Brown, K.A (1998) Gender ilnd sex - what can ancient DNA tell us? Ancient Biomol. 2,3-15 Brown, TA. and Brown, K.A. (] 992) Ancient DNA ilnd the archaeologist Antiquity 66, 10-23. Ca.lin, A (] 994) Seronegative spondylilrthritides. iVied. [III. 22, 145-151 Colin, A ilnd Fries, J.F (] 975) Striking prevalence of ilnkylos­ ing spondylitis in 'healthy' W27 positive males ond femilles. N. Ellg! f. Med. 293, 835-839. Carmody, M'W and Vary, C.PH. (]994) An empirical method for the estimation of nonspecific PCR primer derived genomic reactivity. Biotechni'llIes 16, 1044-1050. de la Chapelle, A., Tippett, P.A, Wetterstrand, G and Pilge, D. (] 984) Genetic evidence of X-Y interchange in a human XX male. Nllture 307, 170-17L Charron, D. (Ed.) (1997) Genetic Diversity of HLA: Functional and Medical Implications. EDK, Paris. Chou, Q., Russell, M" Birch, D.E., Raymond,). ilnd Bloch, W. (1992) Prevention of pre-PCR mis-priming ilnd primer dimerization improves low-copy-number ilmplifications Nue!. Acids Res. 20, 1717-1723. Colson, LB., Bailey, J.F, Vercauteren, M., Sykes, B.C. and Hedges, R.E.M. (1997) The preservation of ancient DNA and bone diagenesis. Ancient Biomol. 1, 109-117. Del Pozzo, G. and Guardiola,J. (1989) Mummy DNA fragment identified. Nature 339, 431-432. Dominguez, 0., Coto, E., Martinez-Naves, E., Choo, S.Y. and L6pez-Larrea, C. (1992) Molecular typing of HLA-B27 alleles. Immunogenet. 36, 277-282. Don, R.H., Cox, P.T, Wainwright, B.J., Baker, K. ilnd Mattick, Js. (1991) 'Touchdown' PCR to circumvent spurious priming during gene amplification. Nue!. Acids Res. 19,4008. Doran, G.H., Dickel, D.N:, Ballinger, WE., Agee, O.F, Laipis, P.J. and Hauswirth, W.W. (1986) Analytical, cellular and molecular analysis of 8,000-year-old human brain tissue from the Wind over archaeological site Nature 223, 803-806. Evison, M.P. (1996) Ancient HLA: a preliminary investigation Ph.D. thesis, University ot Sheffield. Faerman, M., Filon, D., Kahila, G, Greenblatt, c.L., Smith, P and Oppenheim, A (1995) Sex identificiltion of archae­ ological human remains based on amplificahon of the X and Yamelogenin alleles. Gene 167, 327-332. Filon, D., Faerman, M" Smith, P and Oppenheim, A (1995) Sequence analysis reveals a !J-thalassaemia mutation in the DNA of skeletal remains from the archaeological site of Akhziv, Israel. Nature Genet. 9,365-368. Gaensslen, R.E., Berka, K'M" Herrin, G, Mishler, R.E., Sveum, L.K. and Lee, H.c. (1993) Amplification of a genomic sequence in 19th Century human bone DNA. Naturwls­ senschaften 80,80-81. Gardner, M.J. and Altman, D.G. (] 986) Confidence intervals rather than P values: estimation rather than hypothesis testing. Br. Med. f. 292,746-750. Gill, P., Jeffreys,· A.]. and Werrett, D.). (1"985) Forensic application of DNA 'fingerprints'. Nature 318,577-579.

Evison, M.P., Fieller, N.R.J. and Smillie, D.M. (1999). Ancient HLA: a preliminary survey. Ancient Biomolecules, 3(1), 1-28. ISSN 1358-6122

ANCIENT HLA: A PRELIMINARY SURVEY Gill, P., Lygo, j.E., Fow ler, S.j. and Werrett, D.j. (1987) An evaluation of DNA fingerprinting for forensic purposes. Electrophoresis 8, 38-44. Go loubi noff, P , Paabo, S. ilnd Wilson, A:c. (1993) Evol ution of maize inferred from sequence diversity of an Adh2 gene segment from archaeological specimens. Proc. Nat!. Acad. Sci. USA 90, 1997-200 1. GueJlaen, G., Casanovil , M., Bishop, C , Geldwerth, D. , Andre, G., Fell ous, M.'and Weissenbach, J. (1984) Human XX males with Y single-copy DN A fragm ents. Natllre 307, 172-173. Hagelberg, E il nd Clegg, jB. (1 991) Isolation and character­ iSiltio n of DNA from archileologica l bone. Proc. Roy. Soc. Land. Ser. B 24'1, 45-50. Ha gelberg, E., Sykes, B. and Hed ges, R (1989) Ancient bone ON A ilmplified. Nlltllre 342, -1 85. Hilgelberg, E., Be lI, L.S., Allen, 1., Boyde, A, jones, Sj and Cegg, jB (1991) Analysis of ancient bone DNA: techniques an d applica tions. PrOc. Roy. Soc. Lond . Sa. B 333, 399-407. Handt, 0., Hbss, M., Krings, M. and Paabo, S (1994) Ancient DNA: methodological challenges. Experielllw 50, 524--529 Ha nni, C, Laudet, V , Sakk