Key words:HLA-B39; B*3909; HLA-B27; HLA polymorphism;mass spectrometry; peptides; peptide motifs

Acknowledgments:We thank Masafumi Takiguchi (University ofTokyo) for the B*3901 construct, Fernado Roncal(Centro Nacional de Biotecnologıa, Madrid) forhelp in peptide synthesis, and FranciscoGavilanes (Universidad Complutense de Madrid)for amino acid analyses. We also acknowledgethe help of our colleague Fernando Garcıa andthe excellent technical assistance of JuanaBustos, Josefa Gonzalez-Nicolas (ProteinChemistry Department, CBMSO, Madrid), andAlicia Prieto (Centro de InvestigacionesBiologicas, Madrid) in tissue culture, Edmansequencing, and MALDI-TOF analyses,respectively. This work was supported by grantsSAF97/0182 from the Plan Nacional de IπD andPM95-002 from the Spanish Ministry ofEducation, and by an Institutional grant fromthe Fundacion Ramon Areces to the CBMSO.

Received 24 November,accepted for publication 30 December 1999

Copyright c Munksgaard 1999Tissue Antigens . ISSN 0001-2815

Tissue Antigens 1999: 53: 227–236Printed in Denmark . All rights reserved

227

J. Yague The South Amerindian allotype HLA-B*3909M. Ramos has the largest known similarity in peptideJ. VazquezA. Marina specificity and common natural ligands withJ.P. AlbarJ.A. Lopez de Castro HLA-B27

Abstract: HLA-B*3909 has only been found among South Amerindians,and presumably arose locally in these populations. It differs from B*3901 bya single Tyr to Ser change at position 99. To analyze the influence of thispolymorphism on peptide specificity, pool sequence analysis and sequencingof multiple individual ligands from B*3901 and B*3909 were carried out.Both allotypes bind peptides with Arg2 or His2 and nonpolar C-terminalresidues. However, whereas His2 is the predominant B*3901 motif, a ma-jority of the B*3909-bound peptides have Arg2. In addition, B*3909 bindspeptides with Pro2, and also shows an increased preference for Pro3. Inspite of their differences, both subtypes bind overlapping peptide repertoires,as indicated by the identification of several identical ligands from theirrespective peptide pools. B*3909 is significantly more similar in its peptidespecificity to HLA-B27 than B*3901. This is due to the increased preferenceof B*3909 for Arg2 and to low suitability of HLA-B27 for His2. The simi-larity between HLA-B27 and B*3909 was confirmed by identification of anatural ligand common to both allotypes. In addition, multiple HLA-B27ligands bound efficiently B*3909 in vitro. The results indicate that, amongthe HLA class I allotypes of known peptide specificity, B*3909 is the mostsimilar in its peptide binding properties to HLA-B27, which is absent inSouth Amerindians. This may have implications for the susceptibility ofindividuals in these populations to spondyloarthropathies.

Indigenous populations in South America exhibit limited repertoires

of HLA class I alleles. However, the B-locus appears to have sub-

stantially diversified after the ancestor population settled on the

South American continent, giving rise to many HLA-B alleles not

found elsewhere, whose origin is explained mainly by microconver-

sion events involving ancestor alleles frequent in these populations

(1–5). Two lines of evidence suggest that the HLA-B repertoires of

South American Indians have evolved as an adaptive response to

the local environment. First, the size of the HLA-B repertoire in

unmixed South American populations is not significantly different

from those in North America, but the latter populations show little

Authors’ affiliations:

J. Yague1,M. Ramos1,J. Vazquez1,A. Marina1,J.P. Albar2,J.A. Lopez de Castro1

1Centro de BiologıaMolecular Severo Ochoa,Consejo Superior deInvestigaciones Cientıficas(CSIC) and UniversidadAutonoma de Madrid,Facultad de Ciencias,Cantoblanco, Madrid,Spain,

2Immunology and OncologyDepartment, Pharmacia-CSIC, Centro Nacional deBiotecnologıa, Cantoblanco,Madrid, Spain

Correspondence to:Dr. Jose A. Lopez de CastroCentro de Biologıa Molecular

Severo OchoaUniversidad Autonoma de

MadridFacultad de CienciasCantoblanco28049 MadridSpainFax: π34 91 397 8087

Yague et al : Peptide specificity of B*3909

evidence for class I alleles of recent origin. This suggests that allelic

turnover is taking place in South America, so that new B-locus alleles

progressively replace new ones (3, 5). Second, the new allotypes

usually differ from their ancestors by amino acid changes in the pep-

tide-binding site, and therefore must have distinct peptide binding

and presenting properties (3–5). It is likely that the number of founder

HLA-B alleles in South America was very small. Thus, many of the

new variants derive from a single founder allele and constitute sets of

subtypes with the same serologic specificity. For instance, at least 14

HLA-B35 and 8 HLA-B39 alleles have only been found in South and

Central America (5). Thus, to understand the nature of the adaptive

response of the HLA-B repertoire it is essential to determine how the

newly acquired polymorphism affects peptide specificity.

HLA-B39 has an added interest because it has a B pocket similar

to the ankylosing spondylitis-associated HLA-B27 antigen. HLA-B27

is absent from unmixed South Amerindian populations (6), but

B*3901, which is found in South America and elsewhere, presents

peptides with Arg2 or His2 (7), and binds in vitro a subset of the natu-

ral HLA-B27 ligands (8). Some HLA-B39 subtypes found outside

South America have amino acid changes that affect the B pocket and,

by altering its side chain specificity, totally change the repertoire of

peptides bound to these subtypes. This is the case with B*3902 (7),

also found in South America (B*39022) (5), and B*3910 (9).

B*3909 was originally found in a Warao Indian from the Orinoco

Delta in Venezuela (10), and subsequently in a Xavante Indian from

Brazil (4). More recently it was found at a high frequency among

the Yucpa of Venezuela (39), but has not been found outside South

America. It differs from the B*3901 subtype by a single Tyr to Ser

change at position 99. By its location, this change could affect the

specificity of pockets B and D, which bind residues 2 and 3, respec-

tively (10). In this study, we have determined the peptide motifs of

B*3909, sequenced multiple novel B*3901 ligands, and tested the

binding of HLA-B27-bound peptides to B*3909. The results show

that the peptide specificity of B*3909 is more similar to HLA-B27

than B*3901, identify an individual ligand that is naturally pre-

sented by both B*3909 and HLA-B27, and provide evidence for a

substantial overlap between the HLA-B27- and B*3909-bound pep-

tide repertoires.

Material and methods

DNA, transfectant cell lines and antibodies

HLA-B*3901 genomic DNA cloned into pBluescript (a kind gift of

Dr. M. Takiguchi, University of Tokyo, Tokyo, Japan) plus pSV2neo

and a full-length cDNA clone with the coding sequence of B*3909

228 Tissue Antigens 1999: 53: 227–236

cloned into RSV5neo (10) were separately transfected into the HLA

class I defective HMy2.C1R cells by electroporation as previously

described (11). C1R transfectants were cultured in Dulbecco’s modi-

fied Eagle medium (DMEM) supplemented with 7,5% heat-inacti-

vated fetal bovine serum (FBS) (both from Gibco, Paisley, UK). The

same B*3909 construct was also co-transfected, together with the

human b2-microglobulin gene, into the transporter associated with

antigen processing (TAP)-deficient murine RMA-S cells as follows:

1.5¿107 cells were electroporated on ice with 15 mg of each con-

struct, at 250V and 960 mF. Cells were kept on ice for 10 min both

before and after the electroporation, and were then plated and cul-

tured under selection as previously described (12). The B*2705-

RMA-S transfectant cells used in this study have been described

elsewhere (13). RMA-S transfectants were grown in RPMI 1640 me-

dium containing 25 mM HEPES buffer and 5% heat-inactivated FBS

(Gibco).

The W6/32 mAb, specific for an HLA-A, B, C monomorphic de-

terminant (14), was purified by affinity chromatography with Hi-

Trap protein G, and coupled to CNBr-activated Sepharose 4B (both

from Pharmacia, Uppsala, Sweden) following standard procedures.

The ME1 mAb, specific for HLA-B27πB7πB22 (15), was used as

supernatant of the hybridoma cell culture.

Flow microfluorometry (FMF) analysis

Expression of HLA-B39 or HLA-B27 was measured by FMF using 50

ml of a solution containing 45 mg/ml of W6/32 in PBS, or 50 ml of

culture supernatant of ME-1 respectively, as previously described

(12).

Purification of HLA-B39-bound peptides

About 1010 cells were grown at 37æC in 2,000 ml roller flasks, in

DMEM pH 7.4, containing 7,5% heat-inactivated FBS, 100 mg/ml

Streptomycin sulfate and 100 units/ml Penicillin G (all from Gibco).

B*3901 and B*3909 were purified by affinity chromatography of

the cell lysates obtained in the presence of protease inhibitors, and

their peptide pools were obtained by acid extraction with TFA and

fractionated by high-performance liquid chromatography (HPLC) as

previously described (12, 16).

Edman sequencing

Peptide pools and some individual HPLC fractions were subjected

to Edman degradation, as previously described (16). Residue assign-

ments in the pool sequence experiments were done as described

elsewhere (17).

Yague et al : Peptide specificity of B*3909

Synthetic peptides

Peptides were synthesized in an AMS 422 Multiple Peptide Syn-

thesizer (Abimed, Langelfeld, Germany) using the Fmoc chemis-

try, and purified by reverse-phase HPLC. Their correct compo-

sition was determined by amino acid analysis and mass spec-

trometry as previously described (12). Peptides were stored at 4æC

as stock solutions in water. One peptide (AHDETEFYL) was dis-

solved in aqueous ammonia (0.56% v/v) due to its low solubility

in water.

Mass spectrometry analysis and sequencing

The peptide composition of individual HPLC fractions was analyzed

by matrix-assisted desorption-ionization time of flight (MALDI-

TOF) mass spectrometry using a calibrated Biflex instrument

(Brucker-Franzen Analytik, Bremen, Germany) as previously de-

scribed (18). Peptide sequencing by quadrupole ion trap mass spec-

trometry was carried out using an LCQ instrument (Finnigan MAT,

San Jose, CA, USA). Only one peptide (SRDKTIIMW eluted from

B*3901) was sequenced using the microspray interface as described

by Paradela et al. (18). All other sequences were obtained using the

‘‘nanospray’’ interface originally described by Wilm and Mann (19)

as detailed elsewhere (9, 20). Sequence assignments were usually

confirmed by re-fragmenting (MS3) some ‘‘daughter ions’’ coming

from the parental one (20). Most of the sequence assignments were

further confirmed by reproducing the MS/MS spectrum with the

corresponding synthetic peptides.

Epitope stabilization assay

The quantitative epitope stabilization assay used to assess peptide

binding to HLA in vitro has been described elsewhere (21). Briefly,

RMA-S transfectants were incubated at 26æC for 24 h. They were

then incubated 1 h at 26æC with 10ª4 to 10ª9 M peptides without

FBS, transferred to 37æC and collected for FMF analysis with the

W6/32 (for HLA-B39) or ME1 mAb (for HLA-B27) after 4 h. Binding

was expressed as the C50, which is the molar concentration of the

peptide at 50% of the maximum fluorescence obtained with that

peptide in the concentration range used. Relative binding of

multiple peptides was assessed as follows. First, the C50 of a refer-

ence peptide was calculated. Second, the concentration of each other

peptide required to obtain the fluorescence value at the C50 of the

reference peptide was found by interpolation. This was designated

as EC50, which is equal to the C50 for the reference peptide. Relative

binding was expressed as the ratio between the EC50 values of the

peptides compared. Peptides with EC50Æ5 mM were considered to

229Tissue Antigens 1999: 53: 227–236

bind with high affinity; EC50 values between 5 mM and 50 mM were

considered to reflect intermediate affinity; EC50 between 50 mM and

100 mM indicated low affinity; EC50.100 mM was considered to

reflect lack of binding.

Results

HLA-B39-bound peptide sequencing: B*3909 presents a

natural HLA-B27 ligand

The differences in peptide specificity between B*3909 and B*3901

were determined by pool sequencing of B*3909-bound peptides (Fig.

1), and by sequencing 10 individual ligands from B*3901 (1 heptamer,

1 octamer, 7 nonamers, and 1 decamer) (Fig. 2A) and 12 ligands from

B*3909 (5 octamers and 7 nonamers) (Fig. 2B). The GRYATAA pep-

tide is, to our knowledge, the first heptamer sequenced from a major

histocompatibility complex (MHC) class I-bound peptide pool. The

large majority of these peptides were sequenced by electrospray

quadrupole ion trap mass spectrometry. An example is illustrated in

Fig. 3. The MS3 capability of the LCQ instument was used frequently

to confirm the assignments based on the MS/MS spectra, and to re-

solve ambiguities resulting from incomplete fragmentation of the

peptide by MS/MS. A detailed example of the use of MS3 for sequenc-

ing a B*3901 ligand was previously reported (20).

Three identical peptides were sequenced from both B*3901 and

B*3909 (Fig. 2), indicating that both subtypes bind overlapping pep-

tide repertoires in vivo. One of these peptides (EHAGVISVL) was

also previously sequenced from the B*3801-bound peptide pool (7).

Among the peptides sequenced from B*3909, SRDKTIIMW (Fig.

3) was previously reported from both B*2702 (22) and B*2701 (23).

This result demonstrates that B*3909 binds in vivo a natural ligand

of at least two HLA-B27 subtypes.

Main anchor residues: Arg2 is a predominant B*3909

peptide motif

The main peptidic anchor positions (P) for B*3901 and B*3909, as

in most HLA-B proteins, are P2 and the C-terminal one (PW). Like

B*3901, B*3909 allowed for both Arg and His at P2. However, Arg2

was largely predominant upon pool sequencing: the molar ratio of

pth-Arg relative to pth-His at cycle 2 was 4 for B*3909 (Fig. 1), and

only 2 for B*3901 in a control experiment (data not shown). In

addition, of the 13 B*3901 ligands sequenced, including 10 in this

study and 3 from a previous one (7), 5 (38%) have Arg2 and 8 (62%)

have His2 (Fig. 2A). Among B*3909 ligands (Fig. 2B), 9 (75%) have

Arg2, and only 2 (17%) have His2. These results indicate that

Yague et al : Peptide specificity of B*3909

Fig. 1. Pool sequence analysis of B*3909-bound peptides. Num- positions 2 to 4) or 50% (in positions 5 to 9) over either of the previousbers indicate picomole yields of each pth-amino acid at each cycle of Edman two cycles and their absolute yield was .5% of the total amount of pth-degradation. Residues were assigned as previously described (17). Briefly, amino acids at that cycle. Residues within each group in each cycle area residue was assigned at a given position when its absolute yield in that ordered according to their absolute yields, but this does not necessarily re-cycle was increased at least 50% over either of the two previous cycles, flect their relative frequency in that position. Pool sequencing was donebut not if it decreased compared to the previous one. Assigned residues three times from independent preparations, with similar results. One repre-(underlined) were considered strong when they increased at least 300% (in sentative experiment is shown.

B*3909 polymorphism modulates the relative allowance for Arg and

His at P2, towards increasing preference for Arg2.

Pro2 was also observed in B*3909 upon pool sequencing, where

its pth increased about 400% over the previous cycle (Fig. 1), but

was not reported for B*3901 (7). In our control experiment of

B*3901-bound peptides, pth-Pro at cycle 2 was increased relative to

cycle 1 below 50%, which is the adopted criterion for motif assign-

ments by pool sequencing (data not shown). Thus, our result indi-

cates that preference for Pro2 is increased in B*3909 relative to

B*3901. In addition, one ligand with Pro2 was sequenced from

B*3909 (Fig. 2B).

Only Leu and other aliphatic C-terminal residues were previously

reported for B*3901 (7). We have now identified a B*3901 ligand

with C-terminal Phe, indicating that aromatic PW residues are also

allowed in this subtype. However, a great majority of the B*3901

ligands sequenced in this study had C-terminal Leu (Fig. 2A). Both

Leu and Phe were predominant among B*3909-bound peptides by

pool sequencing (Fig. 1). Among the individual B*3909 ligands se-

quenced, most had C-terminal Leu, one had Met, one had Trp, and

230 Tissue Antigens 1999: 53: 227–236

none had Phe. Thus, B*3909 and B*3901 have very similar C-ter-

minal motifs, consisting predominantly of Leu and, secondarily,

other aliphatic and aromatic residues, but not Tyr.

Secondary anchor motifs: distinct preference of B*3909 for

Pro3

Both B*3901 and B*3909 accommodate a wide variety of residues

at P1. Large aliphatic residues, Asn and Gln were rare or absent

among the sequenced ligands from both subtypes. The only notice-

able difference was an increased frequency of Ser plus Thr residues

in B*3901 (54%) relative to B*3909 (25%).

At position 3, both subtypes accommodated also chemically di-

verse residues. However, a main difference was the higher frequency

of Pro3 in B*3909. This residue was not previously reported for

B*3901 (7), nor was it found among the B*3901 ligands sequenced

in this study (Fig. 2A). In contrast, Pro3 accounted for 18.5% of the

total pth-amino acids detected at cycle 3 upon pool sequencing of

B*3909 peptides (Fig. 1). This value was only 1.7% in our control

Yague et al : Peptide specificity of B*3909

pool sequencing from B*3901 (data not shown). In addition, Pro3

was found in 4 (33%) of the individual B*3909 ligands sequenced

(Fig. 2B). Thus, preference for Pro3 is a feature distinguishing

B*3909 from B*3901. Asp3 was found more frequently among

B*3909 ligands, but it was similarly increased in B*3909 and

B*3901 upon pool sequencing (Fig. 1 and data not shown).

Both subtypes were similar in their relaxed selectivity at P4, P5,

and P6. At P7, no Ala was reported from B*3901 (7), and was found

in only the heptamer from this subtype. In contrast, Ala7 was in 3

(25%) of the B*3909 ligands, which is compatible with an increased

preference for Ala7 in this subtype. Finally, no significant differ-

ences between B*3901 and B*3909 were observed at P8 among pep-

tides longer than octamers.

Peptide binding in vitro: the relationship between B*3909

and B*2705

The similarities between some of the peptide motifs of B*3909,

namely Arg2 and nonpolar C-terminal residues, with those of HLA-

Fig. 2. Amino acid sequence of individual peptides from (A) the trap mass spectrometry, except ARDETEFYL, which was sequenced byB*3901-bound and (B) the B*3909-bound peptide pools. The HPLC Edman degradation. The SRDKTIIMW (see Fig. 3) was sequenced by bothfraction number from which the sequence was obtained, the protein data- methods. Edman sequencing was also used in some cases to confirm as-base match of the sequence, and the molecular mass of the sequenced spe- signments from mass spectrometry, especially those concerning the two N-cies, as determined by quadrupole ion trap mass spectrometry, are indi- terminal peptide residues. Residues with the same molecular mass (I/L andcated. Peptides labelled with asterisk (*) are those that were sequenced from Q/K) were assigned, whenever possible, on the basis of unambiguousboth subtypes. Most of the peptides were sequenced by quadrupole ion matching with database sequences.

231Tissue Antigens 1999: 53: 227–236

B27, and the shared SRDKTIIMW ligand, indicate that both allo-

types have overlapping peptide specificities. To gain further insight

into the extent of this overlap, two issues were addressed by testing

natural HLA-B27 and -B39 ligands, and analogs in which Arg and

His were interchanged at P2, for binding to both B*2705 and

B*3909. The first issue was the relative suitability of Arg2 and His2

for each allotype. Four peptides were tested: a B*3909 ligand with

Arg2 (ARDETEFYL), its analog with His2, a B*3901 ligand with

His2 (YHTDTTVKF), and its analog with Arg2. As shown in Fig.

4A, the two peptides with Arg2 bound to B*2705, whereas those

with His2 did not bind. These results indicate that His2 is not a

suitable residue for B*2705. In contrast, both Arg and His2 were

nearly equivalent for binding to B*3909 (Fig. 4B): both the B*3909

ligand and its His2 analog bound efficiently (C50: 7 mM), whereas

both the B*3901 ligand and its Arg2 analog bound very weakly.

The second issue was whether natural HLA-B27 ligands with

Arg2 would bind to B*3909. To address this, 13 natural HLA-B27

ligands and one B*3909 ligand were tested for binding to B*3909

(Fig. 5 and Table 1). Among the peptides tested, those with C-ter-

Yague et al : Peptide specificity of B*3909

Fig. 3. (A) Sequencing of the SRDKTIIMW peptide from the of carbon monoxide (28 daltons) from ions of the b series. The b4 ion wasB*3909-bound peptide pool by quadrupole ion trap mass spec- generated with very low yield, but was assigned on the basis of its moretrometry. The MS/MS fragmentation spectrum of ion at mass/charge clear detection with the corresponding synthetic peptide. The sequence of(m/z) 575.2 is shown. The assigned peptide sequence is indicated, detailing this peptide was also obtained by Edman degradation, which confirmed thethe major backbone fragment ions according to the nomenclature of Roep- SR and KT sequence assignments (data not shown). (B) MS/MS frag-storff and Fholman (38). Ions of the a series are produced by neutral loss mentation spectrum of the synthetic SRDKTIIMW peptide.

minal Arg, Lys, Ala or Tyr did not bind or bound weakly. This is

consistent with sequence data showing the absence of these motifs

among natural B*3909 ligands (Figs. 1 and 2B). Four peptides with

C-terminal Leu, Val, or Phe bound to B*3909 as efficiently or better

than the B*3909 ligand tested. Thus, besides the common peptide

bound in vivo to both HLA-B27 and B*3909, these results indicate

that B*3909 binds many of the natural HLA-B27 ligands carrying

Arg2 and aliphatic/aromatic C-terminal residues, thus revealing the

basis for an extensive overlap between the B*3909- and the HLA-

B27-bound peptide repertoires.

232 Tissue Antigens 1999: 53: 227–236

Discussion

The contribution of HLA-B diversification among South Amerindi-

ans to help this populations against local antigens should be under-

stood through knowledge of how such polymorphism affects pep-

tide specificity. On the basis of structural homology of the side

chain pockets in the peptide binding site between South American

allotypes and other class I molecules for which the peptide motifs

are known, it has been suggested that the repertoire of HLA-B al-

Yague et al : Peptide specificity of B*3909

leles in South America encompasses almost all the supertypic speci-

ficities, that is, the combination of main peptidic anchor motifs at

P2 and PW of the known HLA-B repertoire (4, 5). Although this

analysis provides a general assessment of peptide specificities, it is

insufficient for three reasons. First, the specificity of a given pocket

can sometimes be influenced by polymorphism outside it through

indirect effects (23). Second, polymorphism that affects residues

located at the boundary of a pocket may alter the structure of the

pocket, and its peptide specificity, in such a way that could be

Fig. 4. Epitope stabilization assay showing the binding of theB*3909 ligand ARDETEFYL, its His2 analog, the B*3901 ligandYHTDTTVKF, and its Arg2 analog to (A) B*2705, and (B)B*3909 on RMA-S cells. Data are means of two individual experi-ments.

233Tissue Antigens 1999: 53: 227–236

difficult to detect by sequence comparisons. Third, polymorphism

in pockets other than those involved in P2 and PW binding may

significantly affect peptide specificity. Thus, comparative analyses

of the peptide binding properties of those HLA-B allotypes orig-

inated in South America are required to understand how they con-

tribute to modulating peptide specificity.

The single amino acid difference (Tyr.Ser99) that distinguishes

B*3909 from B*3901 affects a position that is related to at least the

B and D pockets (24), and could, therefore, affect at least the peptide

motifs at P2 and P3. Indeed, among the main differences between

the two subtypes were those observed at these two positions. Most

noteworthy was the increased preference for Arg2 in B*3909, rela-

tive to B*3901. The results with peptide analogs indicate that Arg2

and His2 are similarly suitable for B*3909. In addition, computer

simulation analyses suggest that the effect of the amino acid change

in this subtype is both to improve Arg2 and to disfavor His2 inter-

action, relative to B*3901 (D. Rognan, personal communication). An-

other important difference between B*3909- and B*3901-bound pep-

tides was the increased frequency of Pro3 in the former subtype.

This suggests that the Ser99 change in B*3909 favors Pro3 in the

D pocket. This residue does not appear to significantly affect the

frequency of other P3 residues, relative to B*3901. Residue 99 is

close to the opening rim of the A pocket (24) and might affect P1

preferences. This would be compatible with the lower frequency of

Fig. 5. Epitope stabilization assay showing the binding of theB*3909 ligand ARDETEFYL and three HLA-B27 ligands toB*3909 on RMA-S cells. The EC50 values of these and other naturalHLA-B27 ligands tested in this assay are given in Table 1. Data are meansof two individual experiments.

Yague et al : Peptide specificity of B*3909

Binding of natural HLA-B27 or B*3909 ligands to B*39091

Peptide Ligand of2 EC50 (mM)

ARDETEFYL B*3909 9

RRYQKSTEL B*2705, 01, 03, 04, 06, 10 10

RRIYDLIEL* B*2705, 02, 04 3

RRWLPAGDA B*2705 50

RRFFPYYV B*2705, 02, 03, 04 5

RRFFPYYVY B*2705, 03, 01 nb

RRARSLSAERY* B*2702 60

GRLTKHTKF B*2705, 01, 02, 04, 10 nb

RRFVNVVPTF B*27053, 02 2

RRSKEITVR B*2705 nb

FRYNGLIHR B*2705 nb

RRIKEIVKK B*2705 nb

RRVKEVVKK B*2705 nb

GRIDKPILK B*2705, 10 nb

1 The peptides were tested for binding to B*3909-RMA-S cells using an epitope stabilization assay.

The B*3909 ligand was used as a reference peptide for calculating EC50 values of the HLA-

B27 ligands (see Material and methods); nb: no binding. Peptides labelled with asterisk (*)

are HLA-B27-restricted Epstein-Barr virus (EBV)-derived epitopes. All other ligands come from

endogenous proteins of the cell2 HLA-B27 or -B39 subtypes from whose peptide pools the ligand was sequenced (16, 18, 22,

23, 33–37)3 Our unpublished results

Table 1

small polar residues (SerπThr) at P1 among B*3909 ligands, rela-

tive to B*3901.

Thus, B*3909 polymorphism does not totally change the reper-

toire of bound peptides relative to B*3901, as happened in B*3902

(7) and B*3910 (9). Indeed, three B*3909 ligands were also found in

B*3901. However, the modulatory effect of the B*3909 change

allows for many peptides to bind differentially both subtypes. One

such example was the B*3901 ligand YHTDTTVKF that failed to

bind B*3909 in vitro.

HLA-B39 is associated with the axial type of psoriatic arthritis

(25) and with progression of early disease (26). In addition, B*3901

is associated with HLA-B27-negative ankylosing spondylitis and

pauciarticular juvenile rheumatoid arthritis in Japanese (27). This

has been correlated with the capacity of B*3901 to present peptides

with Arg2. More recently it has been shown this subtype binds

in vitro a subset of the HLA-B27 ligands (8). The peptide binding

properties of B*3909 make it closer than B*3901 to HLA-B27. This

234 Tissue Antigens 1999: 53: 227–236

similarity is based on the increased preference for Arg2, while main-

taining its specificity for nonpolar C-terminal residues. Indeed,

multiple B*2705 ligands bound efficiently B*3909 in vitro and a

common ligand was sequenced from both B*3909 and two subtypes

of HLA-B27 (22, 23). Thus, B*3909 has the greatest similarity with

HLA-B27 in peptide specificity described so far. HLA-B73, which

also binds peptides with Arg2, is significantly different from HLA-

B27 in the C-terminal peptide motifs (28). One previously reported

B*3901 ligand differs from the peptide found in B*3909 and HLA-

B27 only by its lack of a C-terminal Trp residue (7). Since this

peptide was sequenced by Edman degradation, and its molecular

mass, as determined by mass spectrometry, was not reported, there

is a possibility that a C-terminal Trp residue might have gone un-

noticed. Thus, the same B*3909/B27 common ligand might also be

presented by B*3901. Nevertheless, it is also possible that the

B*3901 ligand is indeed different from that found in B*3909. For

instance, we have recently reported that an octamer and its C-ter-

minally extended nonamer are both bound in vivo to B*2705, but

only the nonamer was found in the B*2701-bound peptide repertoire

(18).

HLA-B27 is absent from unmixed South Amerindian populations

(6). Among the HLA class I allotypes in South America of which the

peptide specificity is known, B*3909 is the closest one to HLA-B27.

It has been suggested that HLA-B27 is a highly efficient restriction

element for viral antigens (29), and is associated to slow progression

of AIDS (30–32). Thus, it might be speculated that the similarity of

B*3909 with HLA-B27 in its peptide specificity might have particular

functional significance among South Amerindians. In addition, it

might endow B*3909 individuals with increased susceptibility to

spondyloarthropathy. However, in spite of significant similarities,

B*3909 differs from HLA-B27 in its much greater preference for

acidic and Pro residues at P3. While this property does not necessarily

impair binding to HLA-B27, as demonstrated by the common B*3909/

B27 ligand, which has Asp3, it suggests that the HLA-B27 and

B*3909 peptide repertoires, although overlapping, are significantly

different. A more detailed comparison of the peptide presenting prop-

erties of these two molecules may be worth pursuing.

In conclusion, the peptide specificity of B*3909 shows that an

HLA-B polymorphism of South Amerindian origin makes the new

allotype closer in its peptide presenting specificity to HLA-B27,

which is absent from these populations but widespread among

other human groups.

Yague et al : Peptide specificity of B*3909

References

1. 11. 21.Belich MP, Madrigal JA, Hildebrand WH et Calvo V, Rojo S, Lopez D, Galocha B, Lopez Galocha B, Lamas JR, Villadangos JA,al. Unusual HLA-B alleles in two tribes of de Castro JA. Structure and diversity of Albar JP, Lopez de Castro JA. Binding ofBrazilian Indians. Nature 1992: 357: 326– HLA-B27-specific T cell epitopes. Analysis peptides naturally presented by HLA-B279. with site-directed mutants mimicking to the differentially disease-associated

2. HLA-B27 subtype polymorphism. JWatkins DI, McAdam SN, Liu X et al. New B*2704 and B*2706 subtypes, and torecombinant HLA-B alleles in a tribe of Immunol 1990: 144: 4038–45. mutants mimicking their polymorphism.South American Amerindians indicate 12. Villadangos JA, Galocha B, Garcia F, Albar Tissue Antigens 1996: 48: 509–18.rapid evolution of MHC class I loci. Nature JP, Lopez de Castro JA. Modulation of 22. Rotzschke O, Falk K, Stevanovic S, Gnau V,1992: 357: 329–33. peptide binding by HLA-B27 Jung G, Rammensee HG. Dominant

3. polymorphism in pockets A and B, andParham P, Arnett KL, Adams EJ et al. aromatic/aliphatic C-terminal anchor inEpisodic evolution and turnover of HLA-B peptide specificity of B*2703. Eur J HLA-B*2702 and B*2705 peptide motifs.in the indigenous human populations of the Immunol 1995: 25: 2370–7. Immunogenetics 1994: 39: 74–7.Americas. Tissue Antigens 1997: 50: 219– 13. Villadangos JA, Galocha B, Lopez de 23. Garcia F, Galocha B, Villadangos JA et al.32. Castro JA. Unusual topology of an HLA-B27 HLA-B27 (B*2701) specificity for peptides

4. allospecific T cell epitope lacking peptideFernandez-Vina MA, Lazaro AM, Marcos lacking Arg2 is determined byCY et al. Dissimilar evolution of B-locus specificity. J Immunol 1994: 152: 2317–23. polymorphism outside the B pocket.versus A-locus and class II loci of the HLA 14. Barnstable CJ, Bodmer WF, Brown G et al. Tissue Antigens 1997: 49: 580–7.region in South American Indian tribes. Production of monoclonal antibodies to 24. Saper MA, Bjorkman PJ, Wiley DC. RefinedTissue Antigens 1997: 50: 233–50. group A erythrocytes, HLA and other structure of the human histocompatibility

5. human cell surface antigens. New tools forCadavid LF, Watkins DI. Heirs of the jaguar antigen HLA-A2 at 2.6 A resolution. J Moland the anaconda: HLA, conquest and genetic analysis. Cell 1978: 14: 9–20. Biol 1991: 219: 277–319.disease in the indigenous populations of 15. Ellis SA, Taylor C, McMichael A. 25. Crivellato E, Zacchi T. HLA-B39 and thethe Americas. Tissue Antigens 1997: 50: Recognition of HLA-B27 and related axial type of psoriatic arthritis. Acta Derm702–11. antigens by a monoclonal antibody. Hum Venereol 1987: 67: 249–50.

6. Immunol 1982: 5: 49–59.Khan MA. Spondyloarthropathies 26. Gladman DD, Farewell VT. The role of[Editorial]. Curr Opin Rheumatol 1998: 10: 16. Garcia F, Marina A, Lopez de Castro JA. HLA antigens as indicators of disease279–81. Lack of carboxyl-terminal tyrosine progression in psoriatic arthritis.

7. distinguishes the B*2706-bound peptideFalk K, Rotzschke O, Takiguchi M et al. Multivariate relative risk model. ArthritisPeptide motifs of HLA-B38 and B39 repertoire from those of B*2704 and other Rheum 1995: 38: 845–50.molecules. Immunogenetics 1995: 41: 162– HLA-B27 subtypes associated to 27. Yamaguchi A, Tsuchiya N, Mitsui H et al.4. ankylosing spondylitis. Tissue Antigens Association of HLA-B39 with HLA-B27-

8. 1997: 49: 215–21.Sobao Y, Tsuchiya N, Takiguchi M, negative ankylosing spondylitis andTokunaga K. Overlapping peptide-binding 17. Falk K, Rotzschke O, Stevanovic S, Jung G, pauciarticular juvenile rheumatoid arthritisspecificities of HLA-B27 and B39. Evidence Rammensee HG. Allele-specific motifs in Japanese patients. Evidence for a role offor a role of peptide supermotif in the revealed by sequencing of self-peptides the peptide-anchoring B pocket. Arthritispathogenesis of spondylarthropathies. eluted from MHC molecules. Nature 1991: Rheum 1995: 38: 1672–7.Arthritis Rheum 1999: 42: 175–81. 351: 290–6. 28. Barber LD, Percival L, Parham P.

9. 18. Characterization of the peptide-bindingYague J, Vazquez J, Lopez de Castro JA. A Paradela A, Garcia-Peydro M, Vazquez J,single amino acid change makes the Rognan D, Lopez de Castro JA. The same specificity of HLA-B*7301. Tissue Antigenspeptide specificity of B*3910 unrelated to natural ligand is involved in allorecognition 1996: 47: 472–7.B*3901 and closer to a group of HLA-B of multiple HLA-B27 subtypes by a single 29. Gomard E, Sitbon M, Toubert A, Begue B,proteins including the malaria-protecting T cell clone: role of peptide and the MHC Levy JP. HLA-B27, a dominant restrictingallotype HLA-B53. Tissue Antigens 1998: molecule in alloreactivity. J Immunol 1998: element in antiviral responses?52: 416–21. 161: 5481–90. Immunogenetics 1984: 20: 197–204.

10. 19. 30.Ramos M, Postigo JM, Vilches C, Layrisse Wilm M, Mann M. Analytical properties of Kaslow RA, Carrington M, Apple R et al.Z, Lopez de Castro JA. Primary structure the nanoelectrospray ion source. Anal Influence of combinations of human majorof a novel HLA-B39 allele (B*3909) from Chem 1996: 68: 1–8. histocompatibility complex genes on thethe Warao Indians of Venezuela. Further 20. Marina A, Garcia MA, Albar JP, Yague J, course of HIV-1 infection. Nat Med 1996: 2:evidence for local HLA-B diversification in Lopez de Castro JA, Vazquez J. High 405–11.South America. Tissue Antigens 1995: 46: sensitivity analysis and sequencing of 31. McNeil AJ, Yap PL, Gore SM et al.401–4. peptides and proteins by quadrupole ion Association of HLA types A1-B8-DR3 and

trap mass spectrometry. J Mass Spectrom B27 with rapid and slow progression of1998: in press. HIV disease. Q J Med 1996: 89: 177–85.

235Tissue Antigens 1999: 53: 227–236

Yague et al : Peptide specificity of B*3909

32. 35. 37.Goulder PJ, Phillips RE, Colbert RA et al. Brooks JM, Murray RJ, Thomas WA, Garcia F, Rognan D, Lamas JR, Marina A,Late escape from an immunodominant Kurilla MG, Rickinson AB. Different HLA- Lopez de Castro JA. An HLA-B27cytotoxic T-lymphocyte response B27 subtypes present the same polymorphism (B*2710) that is critical forassociated with progression to AIDS. Nat immunodominant Epstein-Barr virus T-cell recognition has limited effects onMed 1997: 3: 212–7. peptide. J Exp Med 1993: 178: 879–87. peptide specificity. Tissue Antigens 1998:

33. 36. 58: 1–9.Jardetzky TS, Lane WS, Robinson RA, Lamas JR, Brooks JM, Galocha B,Madden DR, Wiley DC. Identification of self Rickinson AB, Lopez de Castro JA. 38. Roepstorff P, Fohlman J. Proposal for apeptides bound to purified HLA-B27. Relationship between peptide binding and common nomenclature for sequence ionsNature 1991: 353: 326–9. T cell epitope selection: a study with in mass spectra of peptides [Letter].

34. subtypes of HLA-B27. Int Immunol 1998:Boisgerault F, Tieng V, Stolzenberg MC et Biomed Mass Spectrom 1984: 11: 601.al. Differences in endogenous peptides 10: 259–66. 39. Martinez-Arends A, Layrisse Z, Arguello Rpresented by HLA-B*2705 and B*2703 et al. Characterization of the HLA class Iallelic variants: implications for genotypes of a Venezuelan Amerindiansusceptibility to spondylarthropathies. J group by molecular methods. TissueClin Invest 1996: 98: 2764–70. Antigens 1998: 52: 51–6.

236 Tissue Antigens 1999: 53: 227–236