Proc. Nati. Acad. Sci. USA Vol. 91, pp. 6259-6263, July 1994 Immunology A second lineage of mammalian major histocompatibility complex class I genes SEIAMAK BAHRAM*, MAUREEN BRESNAHAN*, DANIEL E. GERAGHTYt, AND THOMAS SPIES* *Diision of Tumor Virology, Dana-Farber Cancer Institute, Harvard Medical School, 44 Binney Street, Boston, MA 02115; and tHuman Immunogenetics Program, Fred Hutchinson Cancer Research Center, 1124 Columbia Street, Seattle, WA 98104 Communicated by Sherman M. Weissman, January 3, 1994

ABSTRACT Major sctibit complex (MHC) polymorphic antigen-presenting molecules (2). The physio- class I genes tpically encode polymorphic peptide-binding logical roles of HLA-E and -F are uncertain (14, 15), but chains which are ubi ly expressed and mediate the HLA-G may have a specific function at the maternal-fetal recognition of intr rtigens by cytotoxic T cells. They interface (16). In addition, the MHC contains a number of constte diverse gene fme in different species and include class I pseudogenes and gene fragments (17); however, all of the numerous cd n acal genes in the mouse H-2 the human class I sequences share close relationships indi- coplex, of which some have been adapted to variously mod- cating their common origin from a typical class I gene. ified hmctions. We have identified a d nt family of five In this report, we describe a family of sequencest in the related sequenes in the human MHC whicb are distntly human MHC which are highly divergent from al of the homologous to dass I i. These MIC genes (MHC class I known MHC class I chains and have presumably been chain-related genes) evolved in parallel with the human class I derived early in the evolution of mammalian class I genes. genes and with those of most Ifnot all ma orders. The The MICA gene (MHC class I chain-related gene A) in this MICA gene in this family Is located near HLA-R and is by far family was identified in a search of the HLA-B region for the most divergent m an MHC dass I gene known. It Is other expressed genes. This work was motivated by the still further by It unusual exon4ntron onition unexplained association of HLA-B27 with rheumatoid and and preerential essin fibroblasts and epithlial clls. inflammatory diseases. The uncommon organization of the However, the pence of disc residues in the MICA MICA gene, its restricted expression, and the characteristic amino acid sequence t ted from cDNA suggess that the amino acid sequence of its product, all indicate that the putative MICA chain fods imirly to typical class I chains putative MICA chain has evolved for afunction that is related and may have the capacity to bind peptide or other short to but is quite distinct from that of typical MHC class I Uganda. These results dine a second lineag of evolutionarily chains.§ conserved MHC cla I -gee. This Implies that MICA and bly other members In this family have been selected for MATERIALS AND METHODS specialied f tios that are either ancient or derived from those of typical MHC class I genes, in analogy to some of the Isolaton of MICA and MICB DNA Sequences. MICA and mouse H-2 gen. MICB cDNA clones were isolated from IMR90 human lung fibroblast (Clontech) and keratinocyte (18) libraries with a Typical class I genes in the major histocompatibility complex total insert probe prepared from cosmid M32A (19) by (MHC) ofall vertebrate species encode membrane-anchored described procedures (19). The closely related MICA and cell surface glycoproteins of about 44 kDa that are nonco- MICB cDNA clones of 1.4 and 2.4 kb, respectively, were valently associated with P2-microglobulin (1, 2). These mol- assigned to their corresponding genomic locations by blot ecules are receptors for intracellular peptides, which they hybridization with the previously cloned and mapped present to cytotoxic T cells with a# antigen receptors, thus cosmids R9A, RSA, K7A, and M32A (19), by comparison of facilitating the immune recognition ofintracellular pathogens cDNA and genomic DNA sequences, and by sequencing and altered self-proteins. Characteristic oftypical MHC class across intron-exonjunctions. Selected amino acid sequences I chains is their polymorphism, which correlates with the were aligned by using the EUGENE program package (Baylor ability of various alleles to bind peptides with different College of Medicine, Houston). sequence motifs (3). At least to some extent, the allelic DNA and RNA Blot Hybiion . Preparations of DNA repertoire ofatypical class I chain represents an evolutionary from human B-cell lines andfrom yeast artificial chromosome record of past pathogen-driven selection (4). This principle (YAC) clones and ofRNA from various human cell lines were may also be accountable for the diversity of MHC class I according to standard protocols (20). All cell lines were from genes among different species and, in several cases, for their the American Type Culture Collection or the Tenth Interna- adptation to specialized functions. But such examples have tional Histocompatibility Workshop (21). Primate DNA sam- so far only been found in the mouse. In addition to the typical ples were a gift from D. Watkins (University of Wisconsin). class I H-2K, -D, and -L genes, the mouse H-2 complex YAC clones Y2, Y3, Y17, Y18, Y19, Y22, and Y23 have been includes numerous of the nonclassical Q., T, and M genes published (22, 23); Y24, Y25, Y36, Y37, Y38, Y67, Y54, and (5-8). Ofthese, some ofthe Tgenes encode ligands forT cells Y53 are from unpublished work (D.E.G.). Blot transfer and with yS antigen receptors (9, 10) and the M3 gene product hybridization of DNA restriction fragments and RNA with specifically presents N-formylated peptides of bacterial and [a-32P~dCTP-labeled MICA cDNA were according to stan- mitochondrial origin (11, 12). dard procedures (20). For Fig. 2, RNA samples of 20 pg per The human MHC contains within about 2 megabases six class I genes (13), of which HLA-A, -B, and -C encode the Abbreviations: MHC, major histocompatibility complex; YAC, yeast artificial chromosome. tThe Human Gene Mapping Workshop symbols for these loci are The publication costs ofthis article were defrayed in part by page charge MICA, MCIB, MICC, MICD, and MICE. payment. This article must therefore be hereby marked "advertisement" *The sequence reported in this paper has been deposited in the in accordance with 18 U.S.C. §1734 solely to indicate this fact. GenBank data base (accession no. L14848). 6259 Downloaded by guest on September 26, 2021 6260 Immunology: Bahram et al. Proc. Natl. Acad. Sci. USA 91 (1994) lane were fractionated in 1% agarose/2.2 M formaldehyde DNA encoding MICB. PCR was carried out using GeneAmp gels before blot transfer. Hybridization and washes were reagents and AmpliTaq DNA polymerase (Perkin-Elmer/ under stringent conditions (19). For Fig. 6A, the final wash Cetus), 0.5 pg oftemplate DNA, and 25 pmol ofeach primer. was in 0.5x standard saline citrate (SSC; ref. 20) with 0.1% Samples were subjected to 40 cycles ofPCR with intervals at NaDodSO4 at 65°C. For Figs. SA and 6B, low-stringency 940C (1 min), 550C (1 min), and 720C (2 min) and the products conditions were employed. DNA blots were hybridized in 5 x were directly cloned into the TA vector pCRII (Invitrogen). SSC containing 50% formamide, 50 mM Tris (pH 7.5), 10 mM Sequences were obtained from double-stranded templates by Na4P2O7, 5 mM EDTA, 0.2%o bovine serum albumin, 0.2% using flanking SP6 and T7 primers (Promega) and several Ficoll, 0.2% polyvinylpyrrolidone, 1% NaDodSO4, 5% dex- oligonucleotides derived from MICA intron sequences. tran sulfate, and sonicated salmon sperm DNA (150 Pg/ml) These were 42 bp 3' of the al exon (3'-GTCTTTTCAATC- for 16 hr at 37°C. Filters were washed in 2x SSC with 0.1% CCCGTC-5'), 39 and 40 bp 5' and 3' of the a2 exon, NaDodSQ4 at 37°C. respectively (5'-TCACTTGGGTGGAAAGGTGAT-3' and Cloning and Sequencing of MICA Agcic Variants. MICA 3'-ACGATCTCAACGGAGTGGAGG-5'), and 40 bp 5' of fragments of 2250 bp including the al, a2, and a3 domains the a3 exon (5'-GTTCCTCTCCCCTCCTTAGA-3'). All of were amplified by the polymerase chain reaction (PCR) from the MICA alleles were verified by sequencing ofat least three genomic DNA of the HLA homozygous typing cell lines clones derived from two independent PCR products. EJ32B, MOU, PITOUT, and YAR (21). The oligonucleotides selected for PCR were derived from the 5' end ofthe al exon (5'-GAGCCCCACAGTCTTCGT-3') and the transmembrane RESULTS AND DISCUSSION sequence (3'-TGTAAGGTACAAAGACGAC-5') of MICA We have used previously cloned cosmids from the 170-kb and were tested with cosmid M32A DNA as a template. With interval between HLA-B and BAT] at the centromeric end of this primer pair, no fragment was amplified from cosmid R9A the MHC class I region (19) to screen various human cDNA A BATI MICB 15 MICA 17 HLA-B _n_m - _ p> - R9A K7A - R5A - M32A 20 kb '-' | FIG. 1. (A) Localization of MICA B and MICB in previously cloned L al a2 a3 TM CY'/3'UT cosmids. From a contiguous series, only the cosmids relevant here- I- I MICA R9A, R5A, K7A, and M32A-are shown; for a restriction map, see ref. 19. Black and shaded boxes on top L CY3'UT 2 kb line correspond to genes; open boxes I i MHC-I labeled 15 and 17 indicate short class I-like gene fragments contained in C Xba I and Pst I restriction fragments EXON 1 of 1.5 and 1.7 kb, respectively. Ar- CACTGCTTGAGCCGCTGAGAGGGTGGCGACGTCGGGGCCATGGGGCTGGGCCCGGTCTTCCTGCTTCTGGCTGGCATCTTCCCT 'TTTGCACCT rows show transcriptional orienta- -23 M G L G P V F L LL A G I F P F A P tions of Exon-intron EXON 2 L genes. (B) orga- 94 CCGGGAGCTGCTGCTGAGCCCCACAGTCTTCGTTATAACCTCACGGTGCTGTCCTGGGATGGATCTGTGCAGTCAGGGTTTCTC 'ACTGAGGTA nization of MICA and comparison -5 P G A A A E P H SL R Y N L T V L S W D G S V Q S GF L T E V ' with the canonical structure of MHC A class I genes (1). Data were derived 187 CATCTGGATGGTCAGCCCTTCCTGCGCTGTGACAGGCAGAAATGCAGGGCAAAGCCCCAGGGACAGTGGGCAGAAGATGTCCTG ;GGAAATAAG from cDNA and 27 H L D G O P F L R O DR Q KO) R A K P 0 G QW A E D V L j mapped cosmid EXON 3 DNA sequences. All splce Junctions 280 ACATGGGACAGAGAGACCAGAGACTTGACAGGGAACGGAAAGGACCTCAGGATGACCCTGGCTCATATCAAGGACCAGAAAGAADDCTTGCAT are between the first and second nu- 58 I W D R E T R D L T G N G K D L R M TL A H I K D Q K E G L H cleotide of a codon. Closed and A - 373 TCCCTCCAGGAGATTAGGGTCTGAGATCCATGAAGACAACAGCACCAGGAGCTCCCAGCATTTCTACTACGATGGGGAGCTC 'TTCCTCTCC shaded boxes indicate exons. L, 89 S L Q E I R V E I H ED S..N T.R S S Q H FYY D G E L F L S leader sequence; al-a3, external do- 466 CAAAACCTGGAGACTAAGGAATGGACAATGCCCCAGTCCTCCAGAGCTCAGACCTTGGCCATGAACGTCAGGAATTTCTTGAAG;GAAGATGCC main sequences; TM, transmembrane 120 Q N L E T K E W T M P Q S S R A Q T L A M N V R N F L K ED A segment; CY, cytoplasmic tail; 3'UT, 3' untranslated sequence. (C) MICA 559 ATGAAGACCAAGACACACTATCACGCTATGCATGCAGACTUCTGCAGGAACTACGGCGATATCTAAAATCCGGCGTAGTCCTG ;AGGAGAACA cDNA sequence (upper lines with 151 M K T K T H Y H A M H A D ) L Q E L R R Y L K S G V V L EXON 4 numbering at left) is of 1382 bp. The 652 GTGCCCCCCATGGTGAATGTCACCCGCAGCGAGGCCTCAGAGGGCAACATTACCGTGACAT2CAGGGCTTCTGGCTTCTATCCC :TGGAATATC polyadenylylation signal is under- 182 V P P M LV V T R S E A S E G N T T V T )R A S G F Y P W A N T lined. Lower lines show the trans- 745 ACACTGAGCTGGCGTCAGGATGGGGTATCTTTGAGCCACGACACCCAGCAGTGGGGGGATGTCCTGCCTGATGGGAATGGAACC :TACCAGACC a3 lated amino acid sequence in the sin- 213 L S W R Q D G V S L S H D T QQW G D V L P D G L.G..T. Y 0 T ge-letter code. Numbering at left re- 838 TGGGTGGCCACCAGGATTTCCAAGGAGAGGAGCAGAGGTTCACCTGCTACATGGAACACAGCGGGAATCACAGCACTCACCCT'GTGCCCTCT fers to the predicted mature protein, 244 W V A T R I CQ G EE O R F T (D Y M E H S G N. . .. T H P V P S - after cleavage of the putative signal EXON T the +1 thus corre- 931 GGGAAAGTGCTGGTGCTTCAGAGTCATTGGCAGACATTCCATGTTTCTGCTGTTGCTGCTGCTGCTATTTTTGTTATTATTATT TTCTATGTC.TCAT peptide;sponds to the firstpositionamino acid in the 275 G K V L V L O S H W 0 T F H V S A V A A AA I F V I I I al A EXON 6 F domain. Intron-exonjunctions (see B 1024 CGTTGTTGTAAGAAGAAAACATCAGCTGCAGAGGGTCCAGAGCTCGTGAGCCTGCAGGTCCTGGATCAACACCCAGTTGGGACG ;AGTGACCAC and Fig. 3) are indicated by solid 306 R C C K K K T S A A E G P E L V S L Q V L D O H P V G T S D H 1 triangles and brackets at the right A 1117 AGGGATGCCACACAGCTCGGATTTCAGCCTCTGATGTCAGATCTTGGGTCCACTGGCTCCACTGAGGGCGCCTAGACTCTACAG ;CCAGGCAGC side. Cysteine residues are circled 337 R D A T O L G F O P L M SD L G S T G S T E G A * J and potential sites for N-linked glyc- 1210 ULAGCACTTA osylation (N-X-S/T) are underlined. 1303 TTTATTGTTGTTGGAGGCTGCAAAATGTTAGTAGATATGAGGCGTTTGCAGCTGTACCATATTA AA See text for further explanations. Downloaded by guest on September 26, 2021 Immunology: Bahram et al. Proc. Nati. Acad. Sci. USA 91 (1994) 6261

1 2 3 f - \ k) Ili

i.~, *-1_

_A. _ - I1 4 01!|

FIG. 2. (Upper) Expression of MICA in various cell lines. The RNA bands ofalmost 1.4 and about 2.4 kb are the MICA and MICB transcripts, respectively. Lane 1, Raji (B cell); lane 2, MOLT-4 (T cell); lane 3, HL-60 (promyelocyte); lane 4, U-937 (monocyte); lanes 5-10, fibroblast and epithelial cell lines (IMR-90, MRC-5, U-373, INT407, HeLa, and L-132, respectively). (Lower) Actin control. FIG. 4. Hypothetical placement of conserved MICA al and a2 libraries for homologous transcribed sequences. The BAT) residues on the ribbon diagram of HLA-A2 (adapted from ref. 30) gene flanking this region encodes a putative ATP-dependent based on the sequence alignment shown in Fig. 3. Major structural nucleic acid helicase unrelated to any other MHC gene deviations ofMICA include a deletion ofamino acid positions 45-49 (unpublished data). MICA and MICB were identified by and an insertion of 6 amino acids at position 147. isolation of cDNA clones corresponding to two distant loca- mic tail (1, 2). As in class I genes (1), these domains and a tions 40 and 110 kb from HLA-B, respectively (Fig. 1A). predicted leader peptide correspond to discrete exons in MICA and MICB are closely related, sharing 91% identity in MICA; however, the leader sequence and the al exon are their coding sequences; the MICB sequence will be recorded spaced by a large intron in MICA, and the cytoplasmic tail elsewhere. and 3' untranslated sequences are fused in a single exon (Fig. The MICA sequence was derived from fibroblast and 1 B and C). The organization of MICA is thus distinct from keratinocyte cDNA clones. The single long open reading all known class I genes. Moreover, also contrasting the frame of 1149 bp extends from a probable translation initia- expression of typical class I genes, MICA RNA was neither tion codon (AUG) at nt 40 to a termination codon at nt 1189 detected in B- and T-cell lines (Fig. 2) nor upregulated by (Fig. 1C). It encodes a polypeptide of 383 aa with a relative v-interferon in HeLa cells (data not shown) but was equally molecular mass of 43 kDa (including the putative leader abundant in various fibroblast and epithelial cell lines (Fig. 2). peptide of 23 aa) (Fig. 1C). Both its sequence and predicted Interestingly, a similar expression pattern is also character- domain structure are similar to those of class I chains, istic of several nonclassical mouse Q and T genes (24, 25). including three external domains (al, a2, and a3), a putative Although MICA shares homology with many sequences in transmembrane segment, and a carboxyl-terminal cytoplas- the immunoglobulin superfamily, it is most distinctively al 10 20 30 40 50 60 70 80 90 MICA EPHSLRYNLTVLSWDGSVQSGFLTEVHLDGQPFLRCDRQKCRAK ..... PQGQWAEDVLGNKTWDRETRDLTGNGKDLRMTLAHIKDQKE (Mhom) ILA cons GS--M--FY-SV-RP-RGEPR-IAVGYV-DTQ-V-F-SDAASPRMEPRA-WIEQEGPEYWDRETQIVKAQSQTDRES--TLRGYYNQSEA (18) ILA-A* 0201 (18) ELA-3*2705 IlI--l--IH-I 1-1 1-1 11 11-11111I-I-IIII 1-1-1111111E1IEI 1l1-1i1l111111 I1IICII IKAI I I ---1I-RI i III (21) ILA-G I1I--l--ISAAI-I l 1-lII lIMI Il-lI l1-l-l-lISICII II II-IVI II II IIEEI IRNTI IHAI IIMN-QIlI lIlIIIIIII (15) R-2Dd I-----IV-AI-Il-FIIIIYMEIIII-NIE-I-I-1IIIENIIYIIIIRIIIIIIIIIEIIIR-AIGNEISFIV ---IA-RIIIIIAG (20) B-2T22b I1-----II-Al-Il-LIIIW-IIIIII-IMIV--ISIKEETIIIA...-ILII-ADNIEQQiRIl-IIGILSERN-MIIVHFIIKIMD (19) * C12 100 110 120 130 140 150 160 170 180 MICA GLHSLQEIRVCEIHEDNST . RSSQHFYYDGELFLSQNLETKEWTMPQSSRAQTLAMNVRNFLKEDAMKTKTHYHAMHADCLQELRRYLKSGV. VLRRT (Mhom) ILA cons -S-T--RMYG-DVGP-GRLL-GYHQYA---KDYIAL-EDLRS--AADTAAQI-QRKW ...... A-RVAEQLRAYLEGT-VEW----- EN-KET-Q-A (26) RLA-A*02 01 -I-IV-I(Il-I IS-WIFI-I II1I---l 1I IIKII II1l--I IIIT-KHI(4) ...... --HIIIIIIIIIII------ll-I-- (25) RLA-B*2705 -I-I--NI Il-II-II IDI--l- 1I1IS--- Ill-Il (SI--I (1111) (-i.. I.-1-111111111I IE-I II----- Ill-lI--l (26) BLA-G Sl-l--WIII-tLIS-I1I1I1-lIEI --- 1l-l1l-l1I11I1--l 1I11I I1SKI IC ...... -I-NI IIRI II Il-llI-H---I1l-lIM-I-I (24) B-2Dd - I---WIA I-I IES-I l-IWI-I---C - I-111-T--I IIMII II C...... -Q-GAI -I I------I-NAI-L-- (30) I-2T22b DI-I--WLQI-IIEI-RHICLWINILI--S-ILPTI-INPSIC-VGN-TVP ...... HISQDIKSH-SDL-QK--IK-IIR-L-S (19) A * A * ar3 190 200 210 220 230 240 250 260 270 MICA VPPMVNVTRSEASEGNITVTCRASGFYPWNITLSWRQDGVSLSHDTQQWGDVLPDGNGTYQTWVATRICQGEEQRFTCYMEHSGNHSTHPVPS %hom) ILK cons D--KTH--HHPI-DHEA-LR-W-L----AE---T-QR--EDQ . TQDTELVETR-A-DR-F-K-A-VVVPS-----Y--HVQ-E-LPKPLTLRW (35) RLA-A*02 01 IA-I IIM-I) AV-I- Il-Il- -l-IS---II l---l-Il--ll . III I-l---l-l-t-l -Q---I--Ill-I-ll II (32) RLA-B*2705 1--11-111 -11 1I- 11----- 1 1---1- 1-111 ------1-111111111111 (35) RLA-G I--l Il--l IVFIYI 11- l-l---- -l I-I-I-I l--IlI IlI IVIII Il-l-l--l-l-l-l II .- --Ill-IlIEIIMIII (34) I-2Dd l--IAI--I IRRP--DV-A1-I-I----ID---I-ILN-IE-.IIEMIIIII I-l--l-l-ISIIIIL-K--KI--I1--l-lIEIIIIII (35) R-2T22b l--IAI ---I IIRP--DV-I1l-i-l----lID---I-ILN-IE-. IIMI II Il-l-l--l-l-l-l1I1IIL-K--SI--I IY-I-I IEI III II (36)

FIG. 3. Distant homology of the MICA amino acid sequence with human and mouse MHC class I chains. In the computer-assisted multiple alignment, the al, a2, and c3 domains of various selected human sequences (HLA consensus, HLA-A*0201, HLA-B*2705, and HLA-G; ref. 2) and mouse sequences (H-2T22b, H-2Db; refs. 10 and 28) are separately compared with the corresponding MICA sequences. Gaps were introduced to maximize homology, which is indicated as percent identity at right. Dashes show identical residues. Vertical lines in the lower five class I sequences identify residues shared with the human class I HLA consensus sequence (second line) (2). This dual representation emphasizes the distinct derivation of MICA. Open triangles below the a2 and a3 sequences indicate the highly conserved cysteine residues, which are at amino acid positions 101, 164, 203, and 259 in typical class I chains (2). Stars below the al and a2 sequences identify conserved residues involved in the tight binding of peptides by class I molecules (29). Downloaded by guest on September 26, 2021 6262 Immunology: Bahram et al. Proc. Nati. Acad. Sci. USA 91 (1994) A A B

-1 1 2 3 4 5 6 7 12 3 4 5 6 -: -<:~ -I v c rT 00 r

_d.o ..

",

s- B HLA -B -C -I -A -G -F I FiG. 6. Evolutionary conservation of MICA. (A) MICA-related MIC .0 bands were detected in Asp 718/Xba Idgested DNA samples from -BI1-A various primates under stringent conditions: human, chimpanzee, I''. Y2 orangutan, baboon, gibbon, marmoset, and tamrin anes 1-7, _ R9A Y37 Y54 Y17 Y19 respectively). (B) At low stringency (see Materials and Methods), _.[7A Y38 Y53 Y18- Y36 specific bands were also detected in HindIll-digested DNA from Y3 Y67 Y22 Y23 goat, pig, cow, dog, and hamster (lanes 1-5), but not in mouse DNA 200 kb (lane 6).

adapted to a modified function, presumably early in the FIG. 5. Identification and mapping of additional sequences re- lated to MICA in the MHC. (A) The MICA cDNA hybridized to five evolution of mammalian MHC class I genes. HindIll restriction fragments in genomic DNA from the CGMI and Comparison ofMICA with other class I chains shows that 721 human B-cell lines under conditions of low stringency (see most of the matching residues are common to all of the Materials and Methods). The upper two bands correspond to MICB aligned sequences (Figs. 3 and 4). Some of these are univer- and MICA, which are contained in the cosmids R9A and K7A, sally conserved in MHC class I chains throughout vertebrate respectively (see Fig. 1A). MICA and another band were missing in evolution, which are extraordinarily diverse (31, 32). Only a DNA from mutant 721.221, in which HLA-A and HLA-B are deleted few residues, 3 in al (His-3, Thr-10, and Asp-29), 4 in a2 (34). The three novel bands were matched by specifically hybridizing (Gln-96, Cys-101, Gly-120, and Cys-164), and 11 in a3 (Pro- fragments in overlapping YACs (lanes Y3-Y36) spanning the entire MHC class I region. (B) Approximate locations of MIC-like bands, 185, Phe-208, Tyr-209, Pro-210, Pro-235, Asp-238, Cys-259, MICC, MICD, and MICE near HLA-E, -A, and -F, respectively, are Val-261, His-263, Leu-266, and Pro-269) have been found to indicated by open boxes below the map ofthe class I region (13, 22, be invariant in typical class I sequences from all vertebrate 23). Lines below the map show the locations of cosmids and YACs. classes and can therefore be considered most critical for the structure of class I molecules (32). Their positions are at related to mammalian MHC class I chains and not to the class intra- and interdomain contacts and in or between #-strands, I-like CD1 (26) and Fc receptor (27) sequences encoded but not in the a-helices involved in peptide binding and T-cell outside the MHC. MICA is about equidistant from various recognition (30, 32). Remarkably, except for four substitu- human and mouse class I chains, with 15-21% and 19-301% tions in the a3 domain (positions 238, 261, 266, and 269), all amino acid sequence identity in the al and a2 domains, ofthese residues are present in MICA, includingthe two pairs respectively, and with 32-36% identity in the immunoglob- of cysteines in the a2 and a3 domains, which form intrado- ulin-like a3 domain (Fig. 3). These degrees of similarity are main disulfide loops (Figs. 3 and 4). Thus, it is likely that the substantially lower than those found among the MHC class I putative MICA chain folds similarly to class I chains, al- chains in any species. For example, the mouse nonclassical though it is highly diverged. Further characteristic of MICA H-2T22 and H-2M3 chains share an average of601% and 67% are its unique transmembrane and cytoplasmic tail se- identity in their al-a3 domains with the H-2K class I chain, quences, 3 extra cysteine residues in the al and a3 domains, respectively (10, 11). All of the human MHC class I chains and several potential N-linked glycosylation sites (Fig. 1C). have diverged to a lesser extent. These relationships indicate Specific residues implicated in the binding of the CD8 co- an independent derivation of MICA, which must have been receptor to class I molecules (33) are absent from MICA. But 1 EPHSLRYNLTVLSWDGSVQSGFLTEVELDGQPFLRCDRQKCRAKPQGQVAEDVLGNKTWDRETRDLTGNGKDLRMTLAHIKDQKE

2 . G .. A * 3 . *A.. * 4 . *A. Y *

5 ...... *A ...... Y ...... * ...... 1 GLHSLQZIRVCEIHEDNSTRSSQHFYYDGELFLSQNLETKEWTNPQSSRAQTIAMNVRNFLKEDAMKTKTHYBAMHADCLQELRRYLKSGVVLRRT

2 ...... - ...... 3.V...... V. E.S. R 4.V...... V. E.S. R

1 VPPKVNVTRSEASEGNITVTCRASGFYPWNITLSWRQDGVSLSEDTQQWGDVLPDGNGTYQTWVATRICQGEEQRFTCYMZESGNBSTHPVPS 2 ...... 3 .... . R... .T. G. 4 ...... R... .T. 5 . *S .RT....T...... FiG. 7. Polymorphism of MICA. Amino acid sequences 1-5 represent five MICA alleles encoding variant residues in the al (Top), a2 (Middle), and a3 (Bottom) domains. Alleles 1 and 5 correspond to the IMR-90 and keratinocyte cDNA sequences, respectively. Alleles 1-4 were identified in the HLA homozygous typing cell lines EJ32B (allele 1), YAR (allele 2), PiTOUT (alleles 3, 4), and MOU (allele 4). Stars indicate silent nucleotide substitutions. Downloaded by guest on September 26, 2021 mnunology: Bahram et aL Proc. Nadl. Acad. Sci. USA 91 (1994) 6263 some ofthe amino acid side chains known to interact with the 1. Hood, L., Steinmetz, M. & Malissen, B. (1983) Annu. Rev. Immu- termini the class I nol. 1, 529-568. of short peptides bound by typical mole- 2. Bjorkman, P. J. & Parham, P. (1990) Annu. Rev. Biochem. 59, cules (29) are preserved (Figs. 3 and 4). Of 4 tyrosine side 253-88. chains (positions 7,59,159, and 171) forminghydrogen bonds 3. Falk, K., Rotzschke, O., Stevanovic, S., Jung, G. & Rammensee, with the amino-terminal residue ofbound peptide, Tyr-7 and H.-G. (1991) Nature (London) 351, 290-296. 4. Hill, A. V. S., Allsopp, C. E. M., Kwiatkowsi, D., Anstey, Tyr-171 are present in MICA. In the putative carboxyl- N. M., Twumasi, P., Rowe, P. A., Bennett, S., Brewster, D., terminal binding pocket, Thr-143 is conserved, but Tyr-84, McMichael, A. J. & Greenwood, B. M. (1991) Nature (London) Lys-146, and Trp-147 are variously substituted. Although the 352, 595-600. characteristic binding ofthe peptide terminus by typical class 5. Weiss, E. H., Golden, L., Fahrner, K., Mellor, A. L., Devlin, J. J., BuIlman, H., Tiddens, H., Bud, H. & Flavell, R. A. (1984) Nature I molecules may thus be altered in MICA, these data suggest (London) 310, 650-655. that the putative MICA chain may have the capacity to 6. Stephan, D., Sun, H., Fischer Lindahl, K., Meyer, E., Hammering, associate with peptide ligands, but it is also possible that a G., Hood, L. & Steinmetz, M. (1986 J. Exp. Med. 163, 1=7-1244. different, nonpeptide moiety is bound. 7. Watts, S., Cmanmer Davis, A., Gaut, B., Wheeler, C., Hill, L. & Goodenow, R. S. (1989) EMBO J. 8, 1749-1759. As the class I gene family has evolved by sequential 8. Wang, C.-R. & Fischer Lindahl, K. (1993) Immunogenetics 38, duplications, we investted whether additional sequences 258-271. related to MICA and MICB are encoded in the MHC. Under 9. Vidovic, D., Roglic, M., McKune, K., Guerder, S., MacKay, C. & conditions oflow stringency, the MICA cDNA hybridized to Dembic, Z. (1989) Nature (London) 346, 646-6. 10. Ito, K., Van Kaer, L., Bonneville, M., Hsu, S., Murphy, D. B. & three fragments in total genomic DNA which were not Tonegawa, S. (1990) CeU 62, 549-561. accounted for by MICA and MICB. By using a series of 11. Wang, C.-R., Loveland, B. E. & Fischer Lindahl, K. (1991) Cell6, overlapping YAC clones (refs. 22 and 23; D.E.G., unpub- 335-345. lished data), these fragments were mapped to distinct loca- 12. Pamer, E. G., Wang, C.-R., Flaherty, L., Fischer Lindahl, K. & Bevan, M. J. (1992) Cell 70, 215-223. tions in the vicinities ofthe HLA-E, -A, and -Fgenes (Fig. 5). 13. Carroll, M. C., Katzman, P., Alicot, E. M., Koller, B. H., Ger- These results demonstrate the existence ofa gene family and aghty, D. E., Orr, H. T., Strominger, J. L. & Spies, T. (1967) Proc. indicate that a primordial MIC gene originated from a pro- Nadl. Acad. Sci. USA 84, 8535-8539. totypic class I gene before the evolution of the class I gene 14. Koler, B. H., Geraghty, D. E., Shimizu, Y., DeMars, R. & Orr, H. T. (1988) J. Immunol. 141, 897-904. family. This is further supported by the conservation of 15. Geraghty, D. E., Wei, X., Orr, H. T. & Koller, B. H. (1990)J. Exp. MIC-related sequences not only inprimates (Fig. 6A) butalso Med. 171, 1-18. in various phylogenetically distant mammalian species (Fig. 16. Kovats, S., Main, E. K., Librach, C., Stubblebine, M., Fisher, S. J. 6B). These characteristics clearly distinguish the MIC gene & DeMars, R. (1990) Science 248, 220-223. 17. Geraghty, D. E., Koller, B. H., Hansen, J. A. & Orr, H. T. (1992) family from the nonclassical mouse H-2Q, -T, and -M genes, J. Immunol. 149, 1934-1946. which are organized in gene clusters that are segregated from 18. Tseng, H. & Green, H. (1992) Proc. Nati. Acad. Sci. USA 89, the classical H-2K, -D, and -L genes (5-8) and exist only in 10311-10315. the mouse and perhaps some closely related rodents. How- 19. Spies, T., Bresnahan, M. & Strominger, J. L. (1989) Proc. Nati. ever, it is peculiar that MICA and the T22b chain, which can Acad. Sci. USA 86, 8955-8958. 20. Sambrook, J., Fritsch, E. F. & haiatis, T. (1989) Molecular be recognized by vdT cells (10), share a similar gap in the al Cloning: A Laboratory Manual (Cold Spring Harbor Lab. Press, domain (Fig. 3); the missing residues are normally associated Plainview, NY), 2nd Ed. with the "45" pocket (Fig. 4), which binds the P2 peptide side 21. Yang, S. Y., Milford, E., Hammei, U. & Dupont, B. (1989) in chain in some but not all typical class I molecules (29, 35). ImmunobiologyofHLA, ed. Dupont, B. (Springer, NewYork), Vol. This common feature could be the result of an independent 1, pp. 11-19. 22. Bronson, S., Pei, J., Taillon-Miller, P., Chomney, M. J., Geraghty, adaptation to a similar function, as no MIC-like sequences D. B. & Chaplin, D. D. (1991) Proc. Natl. Acad. Sci. USA U, were detected in mouse DNA (Fig. 6B), either because of 1676-1680. their divergence or possibly because of loss. 23. Geraghty, D. E., Pei, J., Lipsky, B., Hansen, J. A., Taillon-Miller, Taken together, these results define a second lineage of P., Bronson, S. & Chaplin, D. D. (1992)Proc. Natl. Acad. Sci. USA evolutionarily conserved MHC class I genes. This relation- 8, 2669-2673. 24. Hershberg, R., Eghtesady, P., Sydor, B., Brorson, K., Cheroutre, ship implies a specialized, conceivably ancient role of MIC H., Modlin, R. & Kronenberg, M. (1990) Proc. Nadl. Acad. Sci. gene products in some aspect of antigen presentation and USA 87, 9727-9731. T-cell recognition. This is indirectly supported by prelimi- 25. Wang, Q., Geliebter, J., Tonkonogy, S. & Flaherty, L. (1993) nary data indicating that MICA is polymorphic. A compari- Immunogenetics 38,370-372. son of MICA sequences derived from two different cDNA 26. Martin, L. H., Calabi, F. & Milstein, C. (1966) Proc. Natd. Acad. Sci. USA 83, 9154-9158. libraries andfourHLA homozygous typing cell lines revealed 27. Simister, N. E. & Mostov, K. E. (1989) Nature (London) 337, five different alleles with altogether 18 nucleotide changes in 184-187. the al-a3 domains, of which 14 result in amino acid substi- 28. Sher, B. T., Nairn, R., Coligan, J. E. & Hood, L. E. (1965) Proc. tutions (Fig. 7). These data indicate a potential relevance of Natd. Acad. Sci. USA 82, 1175-1179. MICA to and but their thor- 29. Madden, D. R., Gora, J. C., Strominger, J. L. & Wiley, D. C. autoimmunity transplantation, (1992) Cell 7, 1035-1048. ough interpretation will require exact knowledge of the 30. Bjorkman, P. J., Saper, M. A., Samrai, B., Bennett, W. S., function of the MICA protein. Stringer, J. L. & Wiley, D. C. (1987) Nature (London) 329, 506-512. We thank Dr. D. Madden, Dr. D. C. Wiley, and Dr. K. Fischer 31. Kaufiman, J., Andersen, R., Avila, D., Engberg, J., Lambris, J., indahl for their interest and helpful comments; Dr. D. Watkins for Salomonsen, J., Welinder, K. & Skodt, K. (1992)J. Immunol. 148, DNA samples; and Dr. H. Tseng and Dr. H. Green for the kerati- 1532-1546. nocyte cDNA library. S.B. thank Dr. G. Hauptmann for encour- 32. Grossberger, D. & Parbam, P. (1992) Immunogenetics 36, 166-174. agement and was supported by a long-term fellowship from the 33. Salter, R. D., Bnmin, R. J., Wesley, P. K., Buxton, S. E., la Garrett, T. P. J., Clayberger, C., Krensky, A. M., Norment, Association de Recherche sur Polyarthrite and short-term fellow- A. M., Littman, D. R. & Parham, P. (1990) Nature (London) 345, ships from the Association Claude Bernard and the Philippe Foun- 41-46. dation. This work was supported by an American Cancer Society 34. Shimizu, Y. & DeMars, R. (1989) J. Immunol. 142, 3320-3328. Junior Faculty Reseach Award (D.E.G.), the Cancer Research 35. Dirn, M., Paer, K. C., Shiloach, J., I er, R. V., Tsuchida, nstitute/Elaine R. Shepard Investigator Award (T.S.), and National T., Garfield, M., Biddison, W. E. & Coligan, J. E. (1994) J. Institutes of Health Grants AB31873 (D.E.G.) and AL30581 T.S.). Immunol. 152, 620-631. Downloaded by guest on September 26, 2021