The present invention relates generally lυ a novel protein ryroslne kinase and to genetic sequences encoding same.

Protein tyrosine kinases (PTKs) are structurally well suited to a role intracellular signal transduction. Many growth factor receptors, for example, transduce the extracellular stimulus they receive through interaction with their cognate ligand via an intracellular tyrosine kinase domain. Ai least one nf the nonτβceptor PTKs, namely LCK, is believed to mediate ( e transduction In T- cells of a signal from the interaction of a cell-surface protein (GD4) with a cross-linked anϋ-CD4 antibody.

The broader family of PTKs can be sub-divided on the basis of structural parameters of individual member- ^* . For example, the sre family of PTKs now numbers 8 members (Marth Ci. aL, 1985; Nishizawa d ai, 1986; Semba et al., 1986; Martinez at aL, 1987; Sukcgawa ςt al, 1987; Yamanishi ci αL, 1987; Hrvrrman ~L aL, 1987; D mecki ex aL, 1990), each -with a characteristic oomplement of extra-catalytic domains, including an SH2, an SH3 domain and a variable ligand binding domain. It is clear that a process of gene duplication has taken place in this case, so that the evoludonarily successful thematic structure of this family can be employed in a variety of cellular contexts. Similar PTK structural sub-families exist based around die FGF receptor and the CSF-1 receptor (reviewed in Wilks, 1990).

However, one feature in common ith the aforementioned PTKs is that each kinase bears a single highly related "catalytic" domain.

In accordance with the present invention a protein tyrosine kinase is provided ^■ which is distinct from those previously known. In particular, the protein tyrosine kinase of the present invention Is unique since it pnssmses more than one protein kinase catalytic domain. Furϋicrmore, die kinase does not bear an SH2 domain. The novel protein tyrosine kinase of the present invention represents a new subfamily or class of protein tyrosine kinase.

Accordingly, one aspect of the present invention is directed to an animal protein tyrosine kiuase-like molecule comprising a polypeptide having multiple protein kinase catalytic domains but itu SH2 domain.

Preferably, the polypeptide has two protein kinase catalytic domains.

Preferably, the animal Is a mammal and is most preferably a human or a mouse.

Hereinafter, a protein having these characteristics will be referred to as a ".IA " (from lAnuc Kinase: Janus, in Encyclopaedia Britannic* (lltli Ed) Vυl XV pp 155-156). The present invention is specifically exemplified using JAK1 and JAK2 from humans and mice. 'Ihis is done, however, with the understanding that the present invention extends tn the whole family of JAKs from all animals and to mutants, derivatives, analogues and homnlngnes thereof. The term "protein tyrosine lάnase-like molecule" (abbreviated herein to "PTK-like molecule") is used throughout the specification and claims to emphasise that the present invention encompasses all members of the JAK family and to their mutants, derivatives, analogues and homologues.

in accordance -with the present invention, there is provided a PTK-like molecule. Preferably the molecule is in biological pure or in substantially pure and/or synthetic form. The purity of the preparation is characterised by a Sample comprising at least 70% by weight, preferably at least 80% by weight

and most preferably at least 90% by weight PTK-like molecule. Alternatively, where the purity of the enzyme preparation is not critical, the present invention also encompasses an impure PTK-like molecule preparation but which possesses a substantial amount or JAK activity.

The present invention is directed to a naturally occurring PTK-likc molecule, biologically pure or substantially, pure as hereinbefore defined and to derivatives, functional analogues and homologues thereof. uch derivatives include polypeptide* having single or multiple amino acid substitutions, deletions and/or additions relative to the naturally occurring sequence. These derivatives, functional analogues and homologues also encompass single υr multiple substitutions, deletions and/or additions to any associated molecules such as carbohydrate, lipid and/or proteinacious moieties. Reference herein to "PTK-like molecules" includes all such derivatives, functional analogues and homologues. The present invention also extends to synthetic forms nf the polypeptides which include recombiiiaul molecules and molecules prepared by the stepwise addition of a ino acids to groups of amino acids i defined order.

A range of derivatives and analogues of the PTK-like molecule ore contemplated herein and include altering the molecule at its nucleotide sequence-encoding level during its expression within A ORII or in vitro or post- synthesis modification. Such derivatives and analogues include, but are not limited to, modifications to side chains, incorporation of unnatural amino acids during polypeptide synthesis and the use of crosslinkers and other methods which impose eonformational constraints on the polypeptide or their analogues.

Examples uf side chain modifications contemplated by the present invention include modifications of amino groups such a* by reductive alkylation by reaαion with an aldehyde followed by reduction with aBII^ amidinatiσn with methyiacetimidate; acylation with acetic anhydride; carbamoylation of amino groups with cyanate; trinitrobensylarion of amino groups with 2, , 6,

trinltroben2ene suiphonic acid (TNBS); acylation of amino grσpus with succinic anhydride and tetrahydrophthalic anhydride; and pyridoxylation of ljmne with pyridακal-5'-phospbaic fullowed by reduction with NaBH ₄ .

The guanidino group of orginine residues may be modified by the formation of hererocyclic condensation products with reagents such as 2^-butanedionε, phenylglynxal and glyozal.

Tne carboxyl group may be modified by carbodiimide activation jάa. O* acyϊisourca formtion followed by subsequent deriviiisation, for example, to a corresponding amide.

Sulphyriryl groups may be modified by methods such as carixaymethylation with iodoacetic acid or indnaoetamide; performic acid oxidation to cysteic acid; formation of a mixed disulphides with other thiol compounds; reaction with maleimide, maleic anhydride or other substituted maleimide; formntion of mercurial derivatives using 4-chloromercuribeιuoaic, 4- chloromercuriphenylsulphonic acid, phenyimcrcury chloride, 2-cl loro ercuri-4- nitrophftnol and other mercurials; carbamoylation with cyanatc at alkaline pH.

Tryptophaπ residues may be modified by. for example, oxidation -with N- bromσsucciiiiiiiide υr alkylation of the iπdolc ring with 2-hyriro ^χ y-5-nitrobenzyl bromide or sulpheny) halidcs. Tyrosine residues on the other hand, may he altered by nitration with tctranin-omethanc to form a 3-niιrυtyτosine derivative.

Mnditiration of the i idazole ringe of a histidine residue may be accomplished by alkylation with iodoacetic acid derivatives or N-earbethαxylation with diethylp rocai onalc.

Examples of Incorporating unnatural amino adds and derivatives during polypeptide symhesis include, but are not limited to, use of norleucine 4- amino butyric add, 4-amino-3-hydroxy ^* 5-plιenlpentaι»ic acid, C-amiiiohcuuiυic odd, t-butyiglydne, norvelinc, phcnylglycinc, omithinc, aaroosinc, 4-amino-3 hydroxy^-methylheptanoic add, 2-thienyl alanine and/or D-isomers of amino acids.

Crυsslinkers can be used, for example, to stabilise 3D conformations, using homo-bifunctional oosslinkeis such as the bifuiictiunal imidυ esters having (CH2) _n spacer groups with n-1 to n«6, glutaraldchydc, N hydroxysucάnimidc esters and hetero-bifunctional reagents which usually contain an amino-reactive moiety such as N-hydroxysucdnimide and another group spedfic-reactive moiety such as maleimido or dithio moiety (SH) or carhrκlii ide (t X X )H). In addition, polypeptides could be conformaiionally constrained by. for example, incorporation of C, and N.-met ylamiiiυ acids, iiitrυduiΛiυπ υf double bunds between C, and Cg atoms of amino adds and the formation of cyclic polypeptides or analogues by introdudng cσvalent bonds such as forming an amide bond between the N and C termini, between two side chains or between a side chain and the N or C terminus.

The present invention, therefore, extends to peptides or polypeptides and amino acid and/or chemical analogues thereof corresponding to regions of PTK-like molecules. Preferably, the PTK-Iikc molecules -will retain JAK activity. However, molecules carrying mutations in the catalytic domains rendering these inactive may be useful in, for example, titrating out activity and generation of antibodies such molecules are encompassed by the present invention.

The molecular weights of the PTK-like molecules of the present invention range from 100,000 to 200,000 daltons and preferably from 120,000 to 150,000 daltons.

In a most preferred embodiment, the present inventions provides JAK1 and JAK2. JAKI is an approximately 1142 amino acid molecule with a moleculaj weight Of about 132,000 daltons and a nudeotide sequence shown in Figure 2. JAK2 is an approximately 1.100 amino add molecule with a molecular weight of about 130,000 daltons and with a nudeotide sequence shown in Figure 8.

The present invention is also directed to genetic sequences including DNA, cDNA and m NA which encode the PTK-like molecules hcrcindescribed. Such genetic sequences include single or multiple nudeotide substitutions, ddcϋons and/or additions relative the naturally occurring sequence and extend to sequences encoding die derivatives, functional analogues and homologues of the PTK-like molecules. The present invention also provides these genetic sequences in vector and expression vector systems cither in vi i or in a biological system (Le. eukaryotic or prokaryotic cells) transformed widi such vectors or genetic sequences. In a most preferred embodiment the present invention provides cDNA encoding JAKI nnd JAK2 as set forth in Figures 2 and 8, respectively. A range of mutants can be obtained using standard techniques such as an oligouudcυtide mutagenesis and chemical mutagenftsis, and all such mutants and derivatives are encomμi-uαed by the present invention.

The present invention also provides antibodies to a PTK-like molecule. Such antibodies may be monodonal or polyclonal.

The PTK-like molecule of the present invention have varying utility such as in the phosphoryktion of proteins, incorporation of labels and in the design of analogues, antagonists and agonists of JΛKs.

Accordingly, another aspect of the present invention contemplates a method for phosphorylating a protein comprising contacting said protein with a phosphorylating effective amount of a PTK-like molecule, said molecule comprising a polypeptide having multiple protein kinase catalytic domains but

no SH2 domain for a time and under conditions suffident for said first protein to be phosphorylated. Preferably, the polypeptide has two protein kinase catalytic domains and most preferably is JAKI and/or JAK2 and/or tEeix derivatives.

The present invention is further described by reference to the following non- limiting Figures and Examples.

In the Figures:

Figure 1 is α photographic representation of a Northern analysis of murine and human JAKI.

A. 2μg aliquot* of poly(Λ)+ mRNA from murine tissues: lane 1, lung: lane . 2, liver: lane 3, kidney: lane 4, intestine: lane 5, brain lane 6, skeletal muscle: lane 7, spleen: lane 8, salivary gland: lane 9, placenta: lane 10, mammary gland, were fractionated on a 1.0% agarσse/formaldehyde (Moran ~L aL 1988) gel and the RNA transferred onto a Cenescrccn plus (Dupont) membrane. The transferred RNA was hybridized with a 1.8 kb -"P-labellad murine JAKI probe and die filter autoradiograpbed for lfi hr. at *70°C with two intensifying screens. The relative mobilities of 28S rRNA (upper arrow) and 185 rRNA (lower arrow) are shown.

B. 2μg aliquots of poly(A)+ RNA from the human haemopoictic cell lines: lane 1, IIL60 (lnyelσ-monocytic); lane 2, U937 (monocytic): lane 3, LK63 (pre-B): lane 4, RAJI (B-υell): lane 5, CEM (T-cell): lane 6,

K562 (eiythroleukαemia) were fractionated on a 1.0% agarose/fnrmaldehyde (Moran el al, 1988) gel end d e RNA transferred unto a Cenescreen plus (Dupont) membrane. The transferred RNA was hybridized with a full-length ^S H-Iabelled human JAKI probe and the filter autoradiograpbed for 16 hr. at -70°C with two intensifying screens. The relative mobilities of 28S rRNA (upper arrow) and 18S rRNA (lower arrow) are shown.

- 8 - Figure 2 is a representation showing nudeotide sequence and predicted aminπ add sequence of human JAKI. The DNA sequence is numbered at the end of each fine of sequence from the first nudeotide of the largest done (~~1373), the αmino acid sequence (in one letter code) is numbered from the pfitativc AUG and appears above the line to which it refers. The mo kinase catalytic domains are boxed with arrows, and kinase consensus motifs are enumerated according to the nomenclature of Honks ct ai (1988). The suffix a (e.g. Ha) denotes the kinase related motifs present in the first kmase-rdated domain (designated domain- 1 in Fig.3a) also numbered according to the same nomendaturc. The tyrosine residue in an analogous position to the autophαsphoryiation site of a number of other protein tyrosine kinaseε is marked with an inverted triangle.

Figure 3 is a representation showing: Panel A. Amino-add sequence comparison of the two kinasc-ielatcd domains of JAKI. The amino-add sequences (expressed in one- letter aniiuυ acid code) of the two kinase-related domains (domain- 1 amino-adds 576-825; domain-2 (P I K -domain) amino- adds 868-1130) of JAKI and die human threonlne/serine-speciiϊc kinase CDC2 (24) (amino adds 9-272) arc aligned in order to maximize identity. The kinase-related domains have been divided into three segments and the number of αmino add residues separating eadi segment appears at the end of each line. Motifs hdd in common between at least two of tiiese domains are both bolded and based. Roman numerals above the alignment correspond to the conserved domain nomendature devised by Hanks ~L ai (1988).

Panel B. Hydropathy plot of the human JAKI protein. The protein sequence of human JAKI (induding the 10 extra amino adds which precede the most likdy initiation codon) were analysed by the hydrophilidty algorithm of Kyte and Doolitile (1982) using a

span length of 25 amino adds. The relative locations of the two kinase related domains are marked ta Domain- 1 and PTK. The absence of a hydmphnbic transmembrane domain is clcgrly seeu, as can the presence of a highly hydrophilic region between αmino odds 323 and 350.

Figure 4 is a representation of an analysis of the JAKI protein. Pand A. Cellular proteins of the murine mammary fibroblasi cell line (17) were labelled with -^S-methionine (pond A) and iinmunopredpitated with either pre-i mune (PI) or immune (I) αnti-JAK rabbit antiserum (raised in rabbit M8 against the ρGEX/JΛKl/1 fusion protein or the C-teπninal peptide [M3J) and fractionated on a 9.5% SDS-PAGE gel (Laemmll. 1970). Both rabbit antisera spedfically immunoprcdptaicd an S- labelled pmrein of apparent molecular weight 130.000D.

Pand B. Demonsu atiυn of tyrosine kinase activity in JAKI baαerid fusion proteins, JAKI fusion proteins were generated using pGEX2 (Smidi and Johnson, 1988). The entire domaln-1 region was induded in construct pGEX/JAKl/1. The PTK domain portion of the fusion protein extended to die BαmHI site 15 nudcotldcs 5' of the first glydne nodon of the GXGXXG motif of the ATP binding site. An empty vector control was also performed. The bacteria were induced by the addition of Im IPTG as described by Smith and Johnson (1988) and two 1ml aliquots of the bacteria were removed at 60 minutes and 120 minutes post-induction and lysed -with SDS sample buffer. Western analysis υf die samples was performed using anti-phosphotyrosine antisera (PY-20 [ICN]). The arrow heads mark the positions of the GEX-JAK fusion proteins, iix each induction.

Panel C Construction of the pGEX/JΛK fusion proteins. The locations of the two kinase related domains of JAKI arc shown, and bdow, the structure Of the fusion proteins with the glutαthionc S transferase gene.

Figure 5 is a representation of a sequence comparison between JAKI and JAK2 klnase-related domains. The deduced amino acid sequence of murine JAK2 was compared tn the human JAKI amino acid sequence by application of an alignment programme of the Staden VAX-based suite of Sequence analysis programmes. Asterisks (*) denote identity, dollar signs ($) denote conservative substitutions. Sequences are numbered with respect to the JAKI sequence. The extent of the domain- 1 and PTK domains is shown by arrows above the amino add sequence.

Figure 6 is a graphical representation of a phylogenetic αndysis of the two JAKI Kinasc-like domains. The tree building concept of Fitch and Margoliasl (1967) as implemented by Feng and Doolittle (1987) and Hanks el ai (1988) was used to generate a phylogenetic uee as described in Example 1. In each case the catalytic domain alone was used for comparison. The two kinase rdated domains of the JAKI protein were compared indepeudendy. Branch order is a function of structural similarity, branch length a function of sequence identity. The abbreviations used are: SRC= csrc; YES" c-Ycs; FES* c-fes; CSH-R» -Colony stimulating factor- 1 receptor; KIT* c-kit; PDGF-R* Platelet derived growth factor receptor-A," RET= r-KET; ANP-A= Atrid naturetic peptidβ receptor-A; ANP-B » Atrial naturetic peptirift receptor-B; MOS= c-mos; PBS2= polyxin B antibiotic resistance gene product; STE7= sterile mutant wild-type allele gene product; JAKl/1* Domain-1 of Human JAKI; JAKI /I - PTK domain of Human JAKI.

Figure 7 is a diagramatic representation showing models for the rule of members of the JAK family of PTKs in signal transduction. Two possible scenarios arc considered based on an extrapolation of the current notions of the role of PTKs in signd transduction. In panel A the N-terminal domain of the JAK protein serves to sense a particular metabolic cue and convert this input into two distinct outputs. Presumably the output of the second PTK* related domain is tyrosine kinase activity, the activity of Domain* 1 remains unknown. In panel B an alternative scenario is considered. In this case the function of Domain* 1 is the regulation of the PTK domain. In this scenario the sole output of the JAK piυtciπ is the PTK activity.

Figure 8 is a representation of a nudeotide sequence and predicted amino acid sequence of murine JAK2. The nudeotide sequence is numbered beneath each line of sequence, from the first nudeotide of the mast 5' done. The predicted amino add sequence, in one letter code, is numbered at the end of each line of sequence. The two putative kinase domains are shown boxed with arrows, and die kinase consensus modfs are enumerated according to the nomendature of Hanks ct al (1988). The subscript a denotes the kinase-related motifs present in the first kinαse-relatcd domain, which are numbered according to the same nomendature.

Figure 9 is a photographic representation showing expression of JAK2 in murine tissues. Northern blot analysis of 5 ug of mRNA from each of the tissues shown on top of the figure and from various murine (3UF: mammary fibroblasts; 31A mammary epithelial cells; 30.1: factor independent subline of the hemopoietic cell line FDC P1; N1H: fibroblasts) and human (K562; dironic myulogenous leukaemic mils) cell line. The blots were hybridized with a "P- iabdled 22 kb JAK2 probe and autoradiography was for 4 days. The relative mobilities of the 28S and die 1SS rRNA are indicated.

Figure 10 is a graphical representation showing comparison of JAKI and TYK2 amino add sequences. The amino add sequences of JAKI (Wϊlks et ύ, 1991)

«~ and TYK2 (Fiπnbach-Kraft et al, 1990) were compared using the HOMOLOGY option in the programme SEQMATCH, using a window length of 21 amino adds. The ordinate of the graph represents die percentage identity between the two sequences, the abscissa represents the amino add pσstiou i JAKI at which the particular level of identity was calculated. The shaded boxes bdσw the graph represent arbitrarily ascribed JAK homology domains as discussed in the text and further demonstrated in Figure 11.

Figure 11 is a representation showing amino acid sequence comparison of members of the JAK family of PTKs. The amino add sequences of JAKI (Wife et al. 1991) (designated Jl in this figure), JAK2 (J2 in this figure), and TYK2 (Firmbach-Kraft et al, 1990) (T2 in this figure) were aligned using the CLUSTAL program (Higgins and Sharp. 1988). The numbering system is rdative only to the first amino acid of JAKI. and does not take into account the insertion of gaps into this sequence; it is therefore useful only as a rdative measure of location. The extent of each of the JAK homology domains was determined with reference to the homology plot shown in Figure 10. Amino add positions conserved in at least 2 out of the 3 sequences presented arc bolded and presented below the TYK2 sequence as a consensus sequence.

Figure 12 is α representation showing a comparison of the JH3/JH4 domain region with SH2 domains. The two SH2 domains of GAP (die more N* terminal domain denominated GAP-N (residues 178-269), the more C- teππinaL GAP-C, (residues 348-438) (Trahey et al, 1988), and the SH2 domain of vci-k (residues 248-354) (Mayer et al, 19SS) were compared with the JH3/JH4 of JAKI (residues 425-536) (Wllks et aL 1991), JAK2 (residues 252- 359) (this manuscript) and TYK2 (residues 449*555) (Firmbach-Kraft et aL TWO). Amino acids held in common between the two dasses of sequence are denoted by vertical lines between the two sets of sequences. Conserved

residues hdd in common by members of the same class of domain are bnlded.

EXAMPLE 1 MATERIALS AND METHODS

Screening gf fiPJSA librar s

Several cDNA libraries were screened according to the protocols outlined in Maniatis et d, (1982). cHNA libraries from Murine NFS TPA activated spleen (Qontech cat. ML1018). murine swiss-albino 3T3 fibroblast (Cloutech ca # 1023b). murine balb/c buue marrow (Qontech caL# ML1007), murine swiss-webster hole brain (Qontech cat# ML1002), murine ICR linoleic add activated pleurd mαcrophage (Qontech cat# ML1005b), and hiπnan lst- trimester foetal liver (Qontech cat* HL1005b) were all generated in λgr 1 1. cDNA libraries from murine Balb/c rrstis (Qontech cat.# ML1020b), murine day 10 embryonic neuro-epithellum (Reid ei aL 1990) and human foreskin fibroblast cell line AG1518 (Qaesson-Wc b Ci. aL 1989) were generated in λgtlO. Around 10 ⁶ recombinants of each of these libraries were screened on each occasion.

Library screening was carried out as follows. 'I ^" he N 22 (JAKI) PCR clone was labelled by nick-translation (Maniatis et al, 19S2) and used to screen the murine libraries. A murine cDNA done of 1.8kb was isolated amongst 3 other positives from the neuru-cpiihelial and bone marrow cDNA libraries. Two full- length human JAKI cDNA dones were isolated from the unαmplified human foreskin fibroblast cell-line AG1518 by using the murine cDNA as α probe. Hybridisation was at 650C in fixSSQ 1% SDS; 05% Blotto; 200 μg/ml sonicated and denatured herring sperm UNA After hybridisation, the stringency of the final wash was 0.2xSSC: 0.1%SDS at 650C. Filters were autoradiographed uvcrniglu using Kodak XAR-5 X-ray film.

14

For JAK2, the murine macrophage was sσeened first widi the FD 17 (JAK2)

PCR done, yielding 5 positives, and a -portion of the longest cDNA done isolated and used to screen the remaining cDNA libraries. Hybridisation conditions were as above for JAKI.

Twυ strategies were employed for the sequencing of JAKI and JAK2 cDNA dones. In the case of the human JAKI sequence, the Erase-a-Basc kit (FROMEGA) was employed to generate nested ddetions of the largest EcoRI fragment All of the murine JAK2 sequence data, and the remainder of die human JAKI sequence, was determined using oligonudcotide primers based on previously determined DNA sequence. In each case the sequence information was generated using the dideoxynudeotide chain termination method (Sanger ci aL 1977). All sequence information was determined on both strands.

Northern AnaJjsiis

Poly A + mRNA samples were prepared as elsewhere described elsewhere

(Wϊlfcs and Kurbon, 1988). Aliquois (Ipg) were andyscd by deαrophciresis on a 1% agarose gd coπtωning 2J2M formaldehyde; 20mM MOFS,pH 6.8; ImM EDTA; 5mM sodium acetate, and transferred to Hybond (Amcrsham, cat #RPN303N) or nitrocellulose (Schleicher & SchueIl,KA85, cat #401196) membranes.. Filters were prehybridiscd for 41ιr in 50% formaznide containing 3xSSC; 5xDenhardts; lOmM HEPES pH 7.0; lOOμgjπl 1; poly Q100μg/ml denatured herring sperm DNA; lOμg/ml Ecαii DNA; 0.1 % SDS. and hybridised in the same solution with nick-traadated ³² r-Iabclled murine or human JAKI or JAK2 insert, for 18hr. at 42 ^Q C. Filters were washed to a final stringency of 02xS.SC; 0.1% SDS at 65 ⁿ C before exposure to Kodak XAR-5 X- ray film, with two intensifying screens.

jtifeody Bea entt and Protein Analysis

PoJydond rabbit antisera M7 and M8 were raised against affinity pαrificd pGEX/JAM/i bacterial fusion protein (sec section on kinase assays). Poiydonal antibodies M3 and M4 against the C-tcrωinal peptide

(-TSFQNLIECFEALLKC-) of JAKI were raised in rabbits. Peptide was coupled to Keyhole Limpet Hcamocyanin widi 0.05% gluterddehyde, emulsified In Freunds' complete adjuvant and injected intradcπnally at several sites. The animals were boosted four and seven weeks later with coupled peptide emulsified in Freunds' incomplete adjuvant and bled ten days after the fast injection.

Cells were metabolically labelled with either ³⁵ S-methioninc or ^P- orthophosphate in methionine- or phosphate-free medium containing lOOμCi/ωl and lmCi/ml isotope respectively. RIPA- uffer (20mM Tris, pH7.5 comωning 1% Triton XI 00, 1% Na dcαxychoiate, 0.1% SDS, ImM EDTA, ImM PMSF) extracts were incubated on ice with antiserum and πnmune- complexes isolated using Protein A bearing Stφh lncpt.*ιι;i aurcus bacteria. Proteins -were resolved by SDS'PAGE (Laemmli, 1970) and radioactively labelled bands detected by exposure to X-ray film (Kodak XAR-5). The RIKA buffer for ³² P-fobelJed cells contained in addition 20mM EDTA, lQniM NaF, 100 μM ordiovanednte as phospbatase inhibitors.

Fhosphoamino-acid analysis of excised ^P-labelled bands was carried out exactly as described by Hunter and Sefton (1980) Western blot analysis was performed as described by Towbin et al. (1979) as modified in Ziemiecki el al (1990) using either alkaline phospbatase or * ^-labelled protcin-A as a detection system.

Prαtein ϊPnSC AMlVfi

A variety αf protocols have been tried hi urder to reveal the PTK activity of the JAKI protein. First. wrtraction of murine mammary fibroblasts, Reichmann et al (1989) has been performed in a range of buffers, containing Triton-XlOO nr Nonidet P40 (1.0%) alone, or with added Sodium Deαxycholate (0.5% or 1.0%) or in RIPA buffer (containing 1.0% Triton-XlOO; 1.0% Sodium Deoxychαlatc; 0.1% Sodium Dodecylsulpbate). Ceils have been extracted in the resence or absence of phosphatase inhibitors, such as 20mM EDTA, lO M NaF and lOOuM Na2V04.

After immunoprcdpitatiυn, kinase assays have been performed In a range of ATP concentrations (lOOnM-lOmM) or with carrier-free γ-32P-ATP (Amersham cat #101 fi9) in either 20mM Tris. pH 7.4 or 5UmMM HEPES pll 7.4, with either lOmM Mn ^{"1 +} , Mg ^{+ +} or Zn * ⁺ as divalent cation. Incubations have been performed on ice (15 min), at 250C (15 in). at 30°C (15 min) or at 37°C (2 min) in the presence or absence of the phosphatase Inhibitor Na2V04. Finally, γ-32P-GTP was employed as phosphate donor in lieu of Y-32H-ATP, with no success.

In order to generate the JAKl/glutathione translerase fusion proteins sliuwn in Figure 4, domain-l (from nudeotides 1770*2672 In Fig.2) and the PTK domain (from nudeυύdes 2672-end in Fig 2., thus including 5 extra amino acids beyond the ATP binding glyrine motif) were each fused into the BamHI site of pGEX2. The fusion protein was induced by the addition of ImM IPTG as described elsewhere (Smith and Johnson, 1983) and Western blot analysis performed on an induction time course with the M3 anti-JAKl scrum, and the anti- høsphotyrnsine antiserum (Ka ps and Sefton. 1988). Several sources of αnti-phσsphυtyrαslne antisera were tried. The data in Figure 4b were obtained using a commerdally available monodonal antibody preparation FY-20 (ICN). In control experiments, induction of the insert-less pGEX or pGEX/JAKl fusion protein produced no detectable tyrosine phosphorylαϋon of bacterial

substrates and the -reactivity of die aπti-phosphotyrosine antiserum could be completely abolished by the additional of phenyl phosphate.

Computer Aided Sequence Analysis Amino add sequence comparisons were performed using an alignment programme from the Staden-based suite of programmes on a VAX VMS 52.

The phylogenetic andysis of die two inase-like domains of JAKI was performed using the tree-building concept of Fitch and Margoliash (1967) as implemented by Feng and Doolittlc (1987). The SCORE programme used to construct the difference matrices from which d e trees were derived using the

BORD and DLEN programmes, were dl the gift of Or R Dooiiltle of the

University of California ^• San Diego.

The sequence dignment shown in Figure 11 was assembled using the CLUSTRAL program (Higgins and Sharp, 1988) on a VAX VMS S2 minocomputer. The homology plot shown in Figure 10 was assembled using die HOMOLOGY option of the programme SEQMATCH. Database searches with each of the JAK homolgoy domains was reformed using the PASTA programme, based on the Pearson /Lippman algorithm (Pearson and Lippman. 1988).

RACE Anchor PCR

RACE/ Anchor PCR (Frohman el al, 1990; Loh εLaL. 1990) was performed by α modification of the original protocol. Briefly, 2μg of puly(A+) mRNA is converted to cDNA using an Amersham cDNA synthesis kit (cat No. RPN 1256) and 40 ng. of a JAK2 specific oligomideotide primer (5' TACAC TTTAAATATTTTTCT-n Prior to the addition of the reverse transcriptase, the reaction mixture was healed to 65 ^β C. cDNA synthesis was inititated by the addition of 20 units of reverse transcriptase, and the reaction incubated at 55 ⁿ C for 75 minures. The newly sunthesised cDNA was recovered by passage through a spun sephadex column (Maniatis ct aL, 1982) followed by βthanol predpitation. The mRNA/cDNA heteroduplex was G-TnUcd in 30μl

contaimng 1*K) mM potassium cαcodylate, 30 M ris. (pH72), ImM Co ₂ , 0.1mM DTT. 6mM dOTP and 15 units of TdT (IBI), for 10 minutes at 37°C The reaction was terminated by heating to 65°C for 15 minutes and ifaen diluted to 500 μl with lOmM Tris. HC1 (pH7.5). ImM EDT . For (he RACE/Anchor PCR, lOμl of the tailed cDNA was reconstituted into lOOμl PCK buffer (50mM C1, lOmM Tris. HCI[μH83). USmM MgCI ₂ , αθl% gelatin, 200pM of each dNTP) to this was added 50ng of "poiy-C" ϋligonucleotide primer (S'-CTCGAGTCGACGAATTC^-S ^* ) and 25 units of TAQ polymerase (Cetus). ^' Ihe complementary strand of the cDNA was syntiicsised with une cycle of 95°C (5 minutes), S2°C (5 minutes) and 68 ⁿ C (40 minutes), whereupon 500 ug of the "RACE/Anchor" primer (5*- CIU AGTCGACGAATTC-3') and a nested JAK2 specific primer (5'- CTTσCTTAATACTGACATCA-3') were udded and the reaction mix subjected to 30 cycles of 95°C (1 minute), 52°C (2 minutes) and 68 ^β C (5 minutes). The PCR product was phenol/chloroform extracted, precipitated and rcsuspended in lOOμl of water. 'Ihe amplified material was then inascd. size fractionated on a low-melting temperature agarose gel and cloned into Smal cleaved M13mp8. Plaques were screened by hybridisation -with α JΛK2 cDNA. and positives sequenccd.

EXAMPLE 2 Isolation and DNA sequencing of cDNA clones encoding JAKI

JAKI cDNA was cloned using PCR. Northern analysis (Figure la and b) demonstrated that in both mouse and human tissues and cell lines FD22 (JAKI) was encoded by a single widely expressed 5.4kb mRNA. Human cDNA clones of FD22 (JAKI) were isolated from a human foreskin fibroblast cell line (AG 1518) CDNA library (Claesson-Welsh ci. al, 1989). Two of the 8 primary isolates cloned contained inserts which were candidates for being fulJ- length cDNAs ( - 5.3fcb).

19 -

Thc nudeotide sequence of human JAKI h shown in Figure 2. The 5' end of d e done has stop eodons in all 3 reading frames prior to the putative initiation ATG. Two ATG start endow in frame with the longest open reading frame were found at positions 40 and 76 in the nudeotide sequence shown in Figure 2. The first of these is embedded in a particularly poor "Kozak" consensus sequence (Kozak, 1984) (-TAAATGCAG-), whereas the second matches strongly with the optimal consensus sequence defined by Kozak, namely -GCCATGUCT-. The second ATG is considered to be the initiation codon for this protein, since the first one transgresses one of the strangest correlations found In the sequences which precede initiation codυns. namely the presence of a T residue (in lieu of an A residue) 3 nudeoϋdcs before the ATG sequence. At die 3'eud, an hvframe stop eodon at position 3502 defines the C-terminus of the protein. A large (1.405 kb) 3' untranslated region containing a polyadenylation signal completes the mRNA sequence.

The JAKI coding region Of 3426bp encodes a protein of 1142 amino-acids with a calculated molecular mass of 132,000 daltons. The PTK catalytic eomain is located towards the C-terminus of die JAKI protein (Figure 2). In describing the structural features of this domain we have chosen to adopt the nomendature of Hanks fit al (1988). The putative ATP binding site composed of the motif GLY-X-GLY-X-X-GLY- (subdomain 1) followed by an invariant lysinc residue (subdomain II) is located between amino acid residues 871 and 896 of d e JAKI protein. The core motifs of the PTK catalytic domain (sub- domains VI to DC) are also iu dieir appropriate locations, and are well conserved with respect to their primary sequence and their relationship to each Other. The presence of a tyrosine residue at position 1022 In the JAK.1 protein, 11 residues C-rerminal to sub-domain VII (a similarly placed tyrosine is a site of tyrosine autophosphorylation in v-fjpg; Weinmαstcr et al, 1984) is a consistent feature of members of the PTK family and is considered diagnostic of membership of this dass of kinases. The arginine residue at position 1126 (domain XI) marks the end of the highly conserved regions of the PTK

catalytic domain and the entire catalytic domain of 255 amino adds is appruximatdy 28% (with c-fes Wilks and Kuibon. 1988) to 37% (with T ; Koanan ct aL 1988) Identical to other functionally defined PTKs. Finally, there is a rare variant of the highly conserved subdomain VIII motif (residues 1032-1039), which is believed to lie dose to the active site (Hanks ci aL 1988). The presence of phenylalanine and tyrosine flanking the conserved tryptoμhan in this motif is unique to JAKI and JAK2.

A second protein kinav.-related domain (designated here Domain-1) is located between amino adds 578 and 824, 7 amino adds N-teriniiual to the putative PTK domain. All of die conserved elements of protein kinases arc preserved spatially in this domain. In Figure 2 these elements are numbered with respect to their similarity to the subdomain* of protein kinases described by Hanks ct aL (1988) (with the suffix^, e.g. ITI _a ) and die amino add sequences of the two lanases-related domains of JAKI are compared to cadi other and to human GDC2 (Lee aud Nurse. 1987) in Figure 3a. The overall structural similarity of this domain to the kinase domains of both the PTK and threonine/serine kinase families strongly suggest that this region of the protein also functions as a protein kinase. There are, however. Significant differences in the sequences of key motifs within this domain which suggest that Domain-1 may confer o catalytic activity other than serine/thrconine or tyrosine phαsphorylaϋσn. For example, sub-domain Vl _a is poorly conserved with respect to the equivalent motifs in the other kinase families, and the normally invariant -ASP-PHE- GLY- sequence of the PTK and threonine/serine kinase families (sub-domain VU _a ) is replaced by ώe motif ASP-PRO-GL Y- in Domain-1 of JAKI. As has been noted elsewhere, the conservation of the precise sequence of sub-domain VI in the PTK. and threonine/scrin kinase families appears to correlate with the substrate specificity of the kinase. Thus, it is possible that Domain-1 of the JAKI kinase has a substrate spe ifidty other than that exhibited by the PTK and threonine/serine kinase has a substrate specifidty other than thai exhibited by the PTK and thieoiiine/serine kinases. In support of this notion there are subtle differences in the normally consistent spacing between certain

key motifs in Domain-1 of JAKI. The components of die ATP binding site (sub-domains l _a and U _a ) ore some 7 amino adds further apart in this domain that tiiey are in both the PTK family and the threonine/serine kinase. family. Moreover, die spacing beiween sub-domains Vi _a and VII _a in this region is also longer by 9 amino adds. Coiivcrxdy, the distance between sub-domains VII _a and D _a is 7 amino adds shorter than the corresponding region in the PTK catalytic domain. The overall structure of this domain can be expected to be somewhat different to the catalytic domains of the members of the PTK and ϋireonine/serlne kinase families.

The sequences N-tci iual tu Domain- 1 bear no homology to any other portion of a previously described protein kinase. Specifically, no homology was detected to the SH2 domain described for d e cytoμlasmic PTKs such as c- fes/fps. (Sadowski et al 1986) GAP (Trαhey ςt ai 1988) and the phospholipase-C family of proteins (Suh el aL 1988). This is α particularly interesting observation since no other non-receptor PTK has been described which lacks this feature. A hydrophilidty plot failed to demonstrate the present of a hydrophobic domain characteristic of the growth factor receptor type of PTK (Figure 3b) suggesting that this protein is wholly intracellular like other members of the non-receptor dass of PTKs. The one outstanding feature of the JAK hydropathy plot is the highly hydrophilic sequence between residues 320-350. "I is sequence is not conserved in the murine JAK2 protein, however, its remarkable nature suggests that it may well be involved in some function of the JAKI protein.

Expression of JAKI protdn

Several antisera were geneated against the human JAKI protein. Pol donal antisera directed against the hexadeeamer -TSFQNUECFEΛLLKC- (die C terminal 15 amino adds of JAKI) were rairøl in rabbits and used to investigate the nature of die JAKI protein. A second rabbit antiscnim was generated using a pGEX bacterial fusion protein υυniaining the entire ϋomain-1 region of the human JAKI protdn (see Example 1). Preliminary

- 22 - sequence analysis of cDNA dones of murine JAKI demonstrated that the C- teπninus of die human and murine versions of this protein were identical whereas the murine and human Domain-1 regions exhibited a vary high degree of identity. Both systems have thus been used interchangabiy in the " 5 investigation of the properties of the JAKI protdn.

Both antisera have been used for Western blot analyses and immunuprecipitatlon Studies and the data confirm the mRNA expression studies shown in Figure 1. For example, antisera M3 and M8 both 0 immunopredpitate a protein of the same apparent -molecular eight (130 kDaltonε) from ^S-mcthionine labelled murine breast fibroblasts (Fig. 4a). From the same source, ³² P-orthophosphate labelled JAKI was iπununoprecipirated as a phosphothreonine and phosphoscrine wntaiπing phosμhυprotein. It Is a feamre characterstic of members of the protein 5 tyrosine kinase family that they are able to accomplish an act of self phosphoryiation in -ai , Intriguingly, despite the high degree of sequence similarity held by the PTK-related sequence of JAKI to the PTK family in general, it was not possible to demonstrate tyrosine kinase catalytic activity in lmmunσprecipitates of this protein from any of the murine CM human sources 0 tested. A wide range of possibilities has been tested in search of suitable conditions for the demonstration nf this activity. These are listed in Example 1. The reason for the lack of activity may lie with a steric effect of the antibody in d e active site of ihe enzyme.

5 In order to determine whether domain 1 or the PTK domain, in isolation, bor catalytic activity, bacterial fusion proteins of each were generated wilh the gluiadiione transferase protein of Kchic w-i fl japnnie im (Smidi and Johnson, 1988) and an attempt was made to demonstrate with the aid of anti- phosphotyrσsinc antibodies (Kainps and Sefton. 1988) the co-ordinate Ω induction of the fusion protein and tyrosine phosphυrylated protein. In this system there is no cross-reactive background of the anti-phosphotyrsine antisenim, since thwe are no tyrosine kinases in bacteria (Fig.4b). The

- 23 - phosphαrylation of bacterial proteins on tyrosine is dius easily detectable with such a serum. In this series of experiments neither pGEX without insert nor pGEX bearing Domain-1 (pGEX/JAK/1/1) demonstrated 017 tyrosine kinase activity. The pGEX/JAK/1 fusion protein was further purified by affinity chromatography on a reduced glutathione column and have failed to detect any kinase activity using cither histones, casein or enolase as an exogenous substrate. The pattern of indudble tyrosine phospliorylation exhibited by the pGEX PTK fusion protein (pGEX/JAK/2) (Fig. 4b) is ususually simple for an ectopiually expressed PTK fusion protein. Remarkably, the autophosphorylutiυπ of the fusion protein itself does not seem to occur, on observation which may go some way toward explaining why we have had difficulty in demonstrating PTK activity in the intact protein.

cDNA done* covering the coding region of die PCR done FD17 (JAK2) have been isolated firom a range of murine cDNA libraries. The predicted aniiuυ add sequences of JAK2 and JAKI show several regions of significant similarity to each odier (Fig. 5, see also Example 3).

Phylogenetic analysis The phylogenetic rdationship of the catalytic domains of most of the protein kinases has been determined using the tree-building programme of Feng and Oooliide (1987). Figure 6 shows the phylogenetic relationship of the two kinase-rdated domains of the JAK1 protein to the rest of the kinase family. It is concluded from this family tree that these two domains had a common ancestor -which predated die devdυpraent of the PTK sub-family. It is of interest to note that the kinase related domains of the ANP-receptor gnsnylate cydase family diverge at a point dose by.

EXAMPLE 3 Cloning and sequencing of JAK2

Sequence of Marine JAK2

The PCR done FD17 was used as a basis to begin the doning of longer cDNA clones of murine JΛK2. cDNAs were isolated from a range of cDNA libraries, and by RACE (Frohman et al, 1989, oh cl al, 1989). The sequence of murine JAK2 is presented in Figure 8. The predicted amino acid sequence indicates that this protein is highly related to JAKL At die C-terminus, and extending approximately 270 amino adds towards die N-terminus (AA 715- 980), are sequences bearing all the hall marks of a PTK catalytic domain. These are labelled in Figure 8 according to the Hanks nomendature. Immediately N-teπnind to this (AA 400*660) ties the kinase-rdated domain characteristic of this dass of PTKs (Wilks el al, 1991). The approach outlined in Example 2 in relation to JAKI was followed and assigned these kinase related domains according to the Hanks nomendature, appending the suffix Nn to denote their origin. One unusual feature of this domain is an apparent insertion of seven amino acids between elements Vila and Villa (Hanks nomendature; Hanks and Quinu, 1991) with respect to other members of this family. This feature appeared in only one done of the four sequenced which covered this region, and it remains possible that its presence Is due to an infrequent splicing abberation, rather than being of functional significance.

Distribution of JAK2

Northern analysis of die expression of JAK2 in the mouse demonstrated two mRNA transcripts (4.8 and 4.4 kb) hybridizing to die JAK2 probe under low and high stringency hybridization conditions (Figure 9). It is Intriguing to note thai the levels of these transcripts alter with respect to one another In different tissues. For example, the kidney, spleen and lung appear to express predominantly die larger form, whereas ovary, placenta, skeletal (sk) musde and all murine cell lines analyzed e press both forms at about equal levels.

Under low stringency hybridization conditions the murine JAK2 probe recognizes human JAK2 RNA (K562), however, only the smaller transcript of 4.4 kb could be detected. At this point, the origins of either of the two transcripts are undear and no differential splicing events whidi could account for the differences between them could be detected. However, the major source of size differential in these transcripts may lie in the use of different poly-adenylation signals. JAK2 is widely expressed in mouse organs, albeit to different levels. High expression was found in thymus, skeletal musde, ovary and placenta, but J AK2 transcripts were barely detectable In testes or liver. In addition, JAK2 expression was detected in murine cell hues of fibrobiastic (30F, NTH), eμidielial (31 ) and emopoietic (30.1) origin.

JAK Family Homology Domains.

The cloning of JAKI and JAK2 has facilitated the identification of JAK family homology domains. Fig 10 shows α comparison of the amino add sequences of JAKI. Sequence identity between these two proteins manifests itself as seven dearly defined homology domains. These seven domains are defined at a primary sequence level In Figure 11. The PTK domain is dessifled as die JAK-homology Domain 1 (JH1). the second kinase related domain as the JI 12 Domain, and so on to JH7. The boundaries of die JAK homology domains arc arbitrary, and may or may not define functional domains. However, their delineation is a useful device to aid the consideration of the overall structural similarity of this da« of proteins. The structure of the JH1 and JH2 Domains arc described in Example 2. 'l ^ή e JH3 is one of the least highly conserved of the JAK homology domains, each family member bearing between 35%

(JAK2) to 50% (JAKI) of die deduced concensus sequence. The JH4 domain bears the sequence -GLYV RWS- dose to its C-terminal boundary, which has some degree of homology to the SH2 domain core sequence (see below). In addition, the most highly conserved sub-domain of this region bears a potential tyrosine phoβphorylatlon site, namely, -VDCYFRI-. Overall, the JH4 domain has between 51% (JAK2) and 64% (JAKI) ot the deduced consensus sequence for this domain. Each of d e remaining JAK homology domains has been

independently screened against the NBKl. and EMBL databases using ihe FASTA programme. There were nυ compelling homologieβ found with anything in these databases. It is conduded diaτ these domains are structurally and functionally conserved in members of the JAK family of PTKs, but may not, in contradistinction to the SH2 and SH3 domains of die a ^~ family of PTKs, have a role to play in other signal transduction molecules.

The apparent absence of an SH2 domain in any of the JAK family of PTKs is intriguing. Subtle sequence similarities have been detected between SH2 consensus sequences and portions of the JFI3 and JH4 domains (H. Hanafusa and A. Bernards, personal communication). Fig 12 shows an alignment of these two domains. Whilst the similarity of the JH3 domain to SH2 domains is mnsr evident in the region surrounding the SH2 core sequence (F VRES). the homology does not extend far in either direction beyond this region, and only reappears again close to the C-terniinal boundary of the SH2 domain. This lade of extensive homology, particularly in many of those elements most highly conserved between SH2 domains (Koch ct al, 1991) (presumably indicating those residues most intimately involved in die conserved fuu ion of this domain), suggests thai die homology detected is either happenstance, or the product of considerable sequence divergence in evolution. The SI 12 domain is currently believed to interact widi phosphorylated tyrosine residues on the substrates of PTKs (reviewed in Pawsou, 1989; Koch et al, 1991). Whether the JH3/JH4 domains play a similar functional role remains to be determined.

EXAMPLE4

To show that JAKs are represented In a range of animals, oligonudcotide probes were prepared and used to amplify and screen genomes from a variety of animals. JAK DNA was detected in Drosoμhila. xenopus. mouse and human genomes. The main conserved sequence was DPG common to all animals tested.

REFERENCES:

Clαesson-Welsh, 1*, Eriksson, A., Westeπnαrk, B. and Hcldin, CH., Erα ώ ^x m- A ^. s-i ^. 11RA ^~ n, 4917*4921. 1989.

Feng, D.F. and Doolittie, RJ ⁷ . Jour. Mni Rwnhirinn 25: 351-360, 1987.

Fitch, W.M. and Margoliash, E., Science 12: 279-284. 1967.

Hunter, T., and Seflun. B M. Prnr N-t A--Λ Sri 77- 1311-1315, 1980.

Ka ps, M.P., and Sefton, D.M Onm-fene Z. 305-315, 1988.

Kozak. M. Nuclrir Acids R«. 12; 857-872, 1984.

Kozma, S.C. Redmond, S.M.S., Xiano-Chang, F., Saurer, ~M. Groner, B„ and Hynes, N.E. EMfiΩ- l 147-154, 1988.

Kyte, J. and Doolittie, R.F. J. Mol. Bir.1 \∑t 105-132, 1982.

Lae ⁱ wnli. T ΓK waniw ( n-dn-) -11: 680685, 1970.

Lee, M.G. and Nurse. P. ^~~ * ^~~~~ (i --Λ--) ^~ n _. 31-35, 1987.

, Ε.T.. and Sambrook, J.. in Cold Spring Harbυr, N.Y. 1982.

Moran, M F., Koch. CA., Sadowski, I., and Pβwson, T. Oπtngf.nr 3: 665-672, 1988.

Reichmann. E„ Ball, R., Groner, B., and Friis, K ~.. j- f ll Rirt|, UJg; 1127- 1138, 1989.

Smith, D ~. and Johnson, K.S. Gene, fit 31-40, 1988.

Suh. P., Ryu, S.H., Moon, K.H., Suh, HW., and Rhee, S.C. Cell 34; 161-169, 1988.

Towbin, H., SleheUn. T., and Gordon, J Pr-c N-t. A**d. Sd. TKA 76; 4350- 4354, 1979.

Weinmaster, G, Zollcr, M.M, Smith. M.. Hinze, H., and Pawson, T. Cell 37 559-568, 1984.

Wilks, A-F. and Kurban, R.K. n-c--- ^~ - 3- 289-294, 1988.

Ziemiecki, A., Mueller, R.G„ Xiao-Chang, F., Hynes, N.E. and Koziua, S.. E EQJ 2; 191-196, 1990.

Dymeckl, S.M., Neiderhuber, J.E, and Desideriυ. S.v. Sdence.24Z: 332-336, 1990.

Firmboch-Kraft, I., Byers, M„ Showes, T., Dalla-Faverα, R^ and Krulewskl, JJ., a------- ~ 1329-1336, 1990.

Frohman. MA, Dush, M.K. and Martin, G.. Proc. Nar Ar- ^{ή ~} -i T A 85 8998-9002, 1988.

Hanks, SJ . and Quiuπ, A.M. M ^p rh tc in ~ ^~ --yt-nin-v ~Q ^~ . 38-62, 1991.

- 29 -

Han s, S.K., Quinn, AM. and Hunicr, T. Sckiur 241: 42-52, 1988.

Higgins, D.G. and Sharp. P.M. Ucne 23; 237-244, 1988.

Hnlrsnan, DA, Cook, W.D. and Dunn. A.R. Hnv Natl. Aead g- us A ~ 8325-8329, 1987.

Koch, CA, Andersnn, D Moran, MJ ⁷ ., Ellis, C, and Pawson, T., 252: 668- 674, 1991.

Loh, E.Y., Elliott, Ji., Cwitla. S., anier. KL. and Davis, M.M. Science 242 217-220, 1989.

MarUi. J D., Peef, R., Krebc, E.G., and Perimuiier, R.M. Cell i 393-404, 1985.

Martinez, R., MatheyPrevot. B.. Bernards, A. and Baltimore, D. Sώaii-C.221 411-414, 1987.

Mayer, BJ., Hamaguchi, H„ and Hanafusa, H _N Mature 332; 272-274, 1988.

Nishizawa, M.. Semba, K, Yoshida, M.C. Yαmamσto, T., Sasaki. M., and Toyoshima, K. Mu CfilLBio 6; 511-517, 1986.

Pawson, T, QncQgcnti 491-495. 1988.

Pearson, W.R. and Lippman, J. Proc, Natl, Acad. Sci. 85: 2444-2448, 1988.

Reid, H.H., Wilks, A.F.. and Bernard. O., Proc. Nat. Λcad. Sci. fii 1596-1600. 1990.

Sadowski. I. Stone, J.C, and Pawson, T. Mfl QdL-BloL & 4396-4408, 1986.

Sanger, F., Nickleu. S.. and Cnnson, A.R^ rac ~ ^~ -t. Aca Sri TISA 2 : 5463- 5467, 1977.

Semba, K., Nishizawa, M, Myajima, N., Yoshida, M.C., Sukagawa, J., Yamanlshi, Y-, Sasaki, M, Yamamoto, T., aud Toyoshima, , Proe. Natl- ΔcaiLSc Si 5459-5463, 1986.

Sukagawa, J., Semba, K., Yamaπashi, Y., Nishizawa, M., Myajima. N.. Kamamoto, T., and Toyoshima, K.. ~-~ ⁿⁱ V~ ^' ~ ^~i'~~ 7' 41-47, 1987.

Trahey, M, Wong, G., Halenbcck, R., Rubinfeld, B Martin, CΛ., Lαdner, M., Long. CM., Crosier, WJ., Watt, K., Koths, KL, and McCormick, F, Seiqpcc

Wilks, A.F., Proee.« in Grnwth Partnr B-^-rr 9- 97-111, 1990.

Wilks, A.F., Harpur, A., Kurban, RJ . Ralph, S J., Zuercher, G., and Ziemledd, A. MnUeiiiar and Cellular Bioltny 1 - 2057-2065, 1991.

Yamamishi, Y„ Fukushigr., S.I., Semba, I , Sukegawa, J., Miyajima, N., Matsubara, K.L, Yamamoto, T., and toyoshima, K, Molee. Ctsli Binl 7: 237- 243, 1987.