ESTRO GEN SIGNALING IN M ETABO LIC DISEASE;A FUNCTIO NAL GENO M ICS APPRO ACH

Transcription

1 From th e D e partm e ntof Bios cie nce s and Nutrition K arolins k a Ins titute t, Stock h olm, Sw e de n ESTRO GEN SIGNALING IN M ETABO LIC DISEASE;A FUNCTIO NAL GENO M ICS APPRO ACH H ui Gao Stock h olm 2006

2 Allpre vious ly publis h e d pape rs w e re re produce d w ith pe rm is s ion from th e publis h e r. Publis h e d and printe d by K arolins k a Unive rs ity Pre s s Box 200, SE Stock h olm, Sw e de n H ui Gao, 2006 ISBN

3 ABSTRACT Es troge ns h ave traditionally be e n as s ociate d w ith fe m ale re production. M ore and m ore e vide nce h as also link e d e s troge ns to m ainte nance of glucos e h om e os tas is and norm al h e art function. M os t of th e k now n e ffe cts of e s troge n are m e diate d via a dire ct inte raction of e s troge n w ith e s troge n re ce ptors (ER s ), ER α and ER β, w h ich re gulate th e e xpre s s ion of s pe cific s e ts of ge ne s. Advance s in w h ole ge nom e s e q ue nce de te rm ination for various s pe cie s and th e de ve lopm e nt of h igh th rough put te ch niq ue s, s uch as ge ne e xpre s s ion profiling and ge nom e w ide ide ntification of D NA-binding s ite s for s pe cific trans cription factors (Ch rom atin im m unopre cipitation (Ch IP) com bine d w ith m icroarray, Ch IP-onch ip), allow for th e firs t tim e a com pre h e ns ive ch aracte rization of ER targe t ge ne s and as s ociate d cis -re gulatory ER D NA-binding s ite s. Th e aim of th is th e s is w as to im prove our unde rs tanding of th e m olecular e ve nts involve d in m e diating th e e ffe cts of e s troge n on glucos e h om e os tas is and h e art function. To th is e nd, w e h ave inve s tigate d th e m e ch anis m s be h ind th e e ffe cts of e s troge n on glucos e h om e os tas is us ing ER α k nock out (ER K O ) and ER β k nock out (BER K O ) m ice. Th e abs e nce of ER α, but not ER β, re s ulte d in glucos e intolerance and ins ulin re s is tance in both fe m ale and m ale m ice. Th is w as s h ow n to be due to profound h e patic ins ulin re s is tance in ER K O m ice. Ge ne e xpre s s ion profiling re ve aled up-re gulation of lipoge nic ge ne s in live rs of ER K O m ice and w e h ypoth e s ize d th at th is is a re as on for th e obs e rve d ins ulin re s is tance (Pape r I). To furth e r e xplore th is finding, w e tre ate d ob/ob m ice, a rode nt m ode lfor obe s ity and type II diabe te s, w ith e s tradioland found m ark e dly im prove d glucos e tolerance and ins ulin re s pons e to glucos e. Ge ne e xpre s s ion profiling re ve aled th at h e patic lipoge nic ge ne s w e re de cre as e d in ob/ob m ous e live rs afte r e s tradioltre atm e nt. Th e s e s tudie s s upport th e s tudie s in Pape r I in s ugge s ting a link be tw e e n th e ins ulin s e ns itizing e ffe cts of e s troge n and de cre as e d e xpre s s ion of lipoge nic ge ne s in th e live r. W e de m ons trate d th at Stat3 is a dire ct targe t ge ne of ER α in m ous e live r and h ypoth e s ize d th at e s troge nic m odulation of lipid m e tabolis m in th e live r, leading to im prove d glucos e tolerance, is m e diate d by Stat3 (Pape r II). To link e s troge n induce d ch ange s in ge ne e xpre s s ion profiles to D NA-binding by ER s, w e e s tablis h e d ge nom e -w ide analys is of prom ote r occupancy by ER α for m ous e live r tis s ue. Th is analys is confirm e d binding of ER α to prom ote r re gions of k now n e s troge n targe t ge ne s. W e s h ow e d th at th e e s troge n re s pons e e lem e nt (ER E) is th e one of th e m os t com m on m otifs pre s e nt in prom ote rs w h e re h ave re cruitm e nt of ER α. Ge ne s link e d to th e ide ntifie d ER α binding s ite s include d im portant ce llular s ignaling m olecules s uch as Stat3 and s e ve ralge ne s involve d in glucos e and lipid m e tabolis m (Pape r III). W e also e m ploye d ge ne e xpre s s ion profiling to s tudy th e m olecular m e ch anis m of e s troge n action in m ous e h e art. O ve rallve ry fe w ge ne s w e re re gulate d in th e h e art com pare d to th e live r. H ow e ve r, w e ide ntifie d Ptgds as an ER β s pe cific targe t ge ne in m ous e h e art and ide ntifie d th e re s pons e e lem e nt th at m e diate s th is s e lective activation (Pape r IV). Inte re s tingly, th e prom ote r of Ptgds w as occupie d by ER α and re gulate d by an ER α s e lective ligand in m ous e live r (Pape r III). Th e ER s e lectivity of Ptgds e xpre s s ion m igh t provide a m ode l to s tudy th e m olecular bas is for re ce ptor s e lectivity in diffe re nttis s ue s. In s um m ary, a functionalge nom ics approach provide s a pow e rfulne w pe rs pe ctive on e s troge n action in diffe re nt tis s ue s and h as th e pote ntialto trans form traditionalre s e arch approach e s to be com e m ore productive and e fficie nt. I

4 LIST O F PUBLICATIO NS I Bryzgalova G*, Gao H *, Ah re n B, Z ie rath JR, Galus k a D, Ste iler TL, D ah lm an-w righ t K, Nilss on S, Gus tafs s on JÅ, Efe ndic S, K h an A. Evide nce th at oe s troge n re ce ptor-alph a plays an im portant role in th e re gulation of glucos e h om e os tas is in m ice : ins ulin s e ns itivity in th e live r. D iabe tologia M ar;49 (3): * Contribute d e q ually II Gao H *, Bryzgalova G*, H e dm an E, K h an A, Efe ndic S, Gus tafs s on JÅ, D ah lm an-w righ t K. Long-te rm adm inis tration of e s tradiol de cre as e s e xpre s s ion of h e patic lipoge nic ge ne s and im prove s ins ulin s e ns itivity in ob/ob m ice : a pos s ible m e ch anis m is th rough dire ct re gulation of s ignaltrans duce r and activator of trans cription 3. M olendocrinol.2006 Jun;20(6): * Contribute d e q ually III Gao H, Fält S, Gus tafs s on JÅ, D ah lm an-w righ t K. Ge nom e -w ide ide ntification of e s troge n re ce ptor alph a binding s ite s in m ous e live r M anus cript IV O ts uk i M, Gao H, D ah lm an-w righ t K, O h lss on C, Eguch i N, Urade Y, Gus tafs s on JÅ. Spe cific re gulation of lipocalin-type pros taglandin D s ynth as e in m ous e h e artby e s troge n re ce ptor be ta. M olendocrinol.2003 Se p;17(9 ): II

5 CO NTENTS III

6 LIST O F ABBREVIATIO NS AR K O AP1 BER K O CH D Ch IP ED L EGF EGP ER ER E ER K O ERT GO IGT Le pr L-PGD S MM NR PM Ptgds SP1 SR EM s Stat3 T2D TSS IV Arom atas e k nock out Activator prote in 1 Es troge n re ce ptor β k nock out Coronary h e artdis e as e Ch rom atin im m unopre cipitation Exte ns or digitorum longus m us cle Epide rm algrow th factor Endoge nous glucos e production Es troge n re ce ptor Es troge n re s pons e e lem e nt Es troge n re ce ptor α k nock out Es troge n re place m e ntth e rapy Ge ne ontology Im paire d glucos e tolerance Le ptin re ce ptor lipocalin-type pros taglandin D s ynth as e M is m atch Nuclear re ce ptor Pe rfe ctm atch lipocalin-type pros taglandin D s ynth as e Sp1 trans cription factor Se lective e s troge n re ce ptor m odulators Signaltrans duce r and activator of trans cription 3 Type 2 diabe te s Trans cription s tarts ite

7 Es troge n s ignaling in m e tabolic dis e as e ;a functionalge nom ics approach 1 INTRO DUCTIO N Es troge ns are th e ve ry firs t is olate d s te roid h orm one s (1). Th e y h ave traditionally be e n conne cte d w ith fe m ale re production (2). H ow e ve r, th e im portance of th e s e h orm one s in th e m ale (3) and for non-re productive proce s s e s, s uch as m e tabolic dis e as e and cardiovas cular dis e as e, h as late r be e n e s tablis h e d (4-6). In th is th e s is, I w illpre s e nt m y s tudie s us ing a ge nom e -w ide and s ys te m atic functionalge nom ics approach to ch aracte rize e s troge n re ce ptor (ER ) targe t ge ne s. Th e ph ys iological focus h as be e n th e e ffe cts of e s troge n in re lation to glucos e h om e os tas is and h e art function. I w ills tart w ith a ge ne ral ins truction to e s troge n s ignaling and w h at is k now n about e s troge n s ignaling in re lation to m e tabolic proce s s e s. I w illth e n introduce th e m ajor te ch nologie s us e d in m y s tudie s follow e d by a pre s e ntation and dis cus s ion of th e obtaine d re s ults. Finally, I w ille nd th is th e s is w ith a ge ne raldis cus s ion and conclus ions. 1.1 H O RM O NES AND H O RM O NE RECEPTO RS Th e e ndocrine s ys te m, along w ith th e ne rvous s ys te m, re gulate s and inte grate s th e function of diffe re ntiating ce lls in allm ultice llular organis m s to m aintain ph ys iologicalh om e os tas is and to coordinate ph ys iologicalproce s s e s s uch as grow th, m e tabolis m, and re production. Th e ne rvous s ys te m acts th rough e lectricalim pulse s and ne urotrans m itte rs to caus e rapid and locale ffe cts. Th e e ndocrine s ys te m acts th rough ch e m icalm e s s e nge rs, h orm one s, to pe rform durable and m ore ge ne ralize d e ffe cts acros s th e w h ole body (7). O n th e bas is of th e ir ch e m icals tructure, th e e ndocrine h orm one s can be clas s ifie d into four ch e m icalcate gorie s : am ino acid de rivate s, polype ptide h orm one s, fatty acid de rivative s and s te roid h orm one s. To m e e t th e clas s icalde finition of a h orm one, th e s e ch e m icals ne e d to be s ynth e s ize d in a s pe cific gland, s e cre te d into th e blood and e xe rt th e ir activity on dis tantce lls or tis s ue s. H orm one s act by binding to s pe cific re ce ptors in th e targe t organ. Th e re ce ptor binds th e h orm one w ith h igh affinity, dis tinguis h ing it from oth e r m olecules, and trans late s th e h orm onals ignaling into an intrace llular re s pons e. H orm one s are carrie d by th e blood th rough out th e e ntire body, but th e y affe ct only ce lls th at e xpre s s th e ir s pe cific re ce ptors. Th is is re m inis ce nt of a lock and k e y m e ch anis m. If th e k e y fits th e lock, only th e n w illth e door ope n. D e pe nde nt on th e ir location in th e ce lls, h orm one re ce ptors are clas s ifie d as e ith e r m e m brane bound re ce ptors or intrace llular re ce ptors. H ydroph ilic h orm one s, s uch as pe ptide h orm one s, cannot cros s th e lipid containing ce llm e m brane and ne e d to inte ract w ith re ce ptors at th e ce lls urface to pe rform th e ir action. Afte r activation, m e m brane bound re ce ptors ge ne rate various k inds of s e cond m e s s e nge rs th at can re gulate th e function of prote ins and th e trans cription of ge ne s th rough diffe re nt m e ch anis m s. H ydroph obic 1

8 H ui Gao h orm one s, s uch as s te roid h orm one s, can cros s ce llm e m brane s and bind to intrace llular re ce ptors. Th e s e intrace llular re ce ptors ofte n act as trans criptionalre gulators to re gulate th e trans cription of s pe cific targe tge ne s (Figure 1). Figure 1. H orm one s w ith diffe re nt ch e m icals tructure s act in diffe re nt w ays. Ste roid h orm one s are ge ne rally s ynth e s ize d from ch oles te rolin th e gonads and adre nal glands (8) (Figure 2). Th e y can be groupe d into: glucocorticoids, m ine ralocorticoids, androge ns, e s troge ns, and proge s tage ns. Vitam in D is also clas s ifie d as a s te roid h orm one alth ough it is a s te rolrath e r th an a s te roid (9 ). Ste roid h orm one s are, w h e n trans porte d in th e bloods tre am, ofte n bound to binding prote ins. Th e y are h ydroph obic m olecules th at can e nte r th e ce llby pas s ive diffus ion th rough th e ce llm e m brane. Afte r e nte ring th e ce ll, th e y bind to and activate s pe cific intrace llular re ce ptors. Th e s te roid re ce ptors be long to th e nuclear re ce ptor (NR ) s upe rfam ily. Th e y m ainly act as ligand-de pe nde nt trans criptional re gulators in re s pons e to activation by th e re s pe ctive ligands. NR s are ch aracte rize d by tw o cons e rve d dom ains ;a D NA-binding dom ain and a ligand-binding dom ain. Afte r activation, s te roid h orm one re ce ptors dim e rize to form h om odim e rs or h e te rodim e rs, w h ich can re cognize s pe cific D NA m otifs nam e d re s pons e e lem e nts in th e prom ote r re gions of targe t ge ne s, to re gulate targe t ge ne trans cription. Th e D NA m otifs re cognize d by s te roid h orm one re ce ptors cons is t of tw o h alf-s ite s s e parate d by 1-5 bas e pairs. Th e dis tance be tw e e n th e h alf-s ite s can vary and th e y can be arrange d as inve rte d or dire ct re pe ats. Th e type of re pe at and th e s pacing be tw e e n th e h alf-s ite s de te rm ine th e re ce ptor s pe cificity of a 2

9 Es troge n s ignaling in m e tabolic dis e as e ;a functionalge nom ics approach give n m otif. Trans cription re gulation by s te roid h orm one re ce ptors also involve s a large num be r of core gulatory prote ins including coactivators and core pre s s ors (10). Figure 2. Sum m ary of s te roid h orm one bios ynth e s is. Ste roid h orm one s are s ynth e s ize d from th e com m on pre curs or ch oles te rol.th e firs t com m itte d s te p is th e conve rs ion of ch oles te rol to pre gne nolone. Es troge ns are m ainly s ynth e s ize d in th e gonads. 1.2 ESTRO GEN SIGNALING Es troge ns Es troge ns, nam e d for th e ir im portance in th e e s trous cycle, are s te roid com pounds and are th e firs t is olate d s te roid h orm one s (1). As th e prim ary fe m ale s e x h orm one, e s troge n prom ote s th e de ve lopm e nt of fe m ale s e condary s e x ch aracte ris tics and is also involve d in re gulating th e m e ns trualcycle. Es troge ns h ave traditionally be e n conne cte d w ith fe m ale re production. Th e im portance of th e s e h orm one s for m an and non-re productive proce s s e s, s uch as m e tabolic dis e as e and cardiovas cular dis e as e, h as be e n e s tablis h e d late r. Es troge ns are prim arily produce d in th e ovary, te s tis and adre nalcorte x, but also by s om e pe riph e ral tis s ue s, s uch as th e live r (11). Th e s e s e condary s ource s of e s troge n are e s pe cially im portant in pos t-m e nopaus al w om e n. Lik e oth e r s te roid h orm one s, e s troge ns are biologically de rive d from ch oles te rol. Th e conve rs ion of ch oles te rolinto pre gne nolone is th e firs t s te p of e s troge ns bios ynth e s is. Finally e s troge ns are produce d by conve rs ion of androge ns (Figure 2). Th e re are th re e m ajor naturally occurring e s troge ns in h um an. Th e y are e s tradiol, e s trioland e s trone (2). Es tradiolis th e m ajor ovarian e s troge n and h as th e 3

10 H ui Gao h igh e s t affinity for both ER s of th e naturale s troge ns (12). In pos t-m e nopaus alw om e n e s trone is th e m ajor e s troge n Es troge n re ce ptors M os t of th e k now n e ffe cts of e s troge n are m e diate d via a dire ct inte raction of e s troge ns w ith e s troge n re ce ptors (ER s ) th at activate th e e xpre s s ion of s pe cific s e ts of targe t ge ne s. Th e firs t ER w as clone d in from a bre as t cance r ce llline (13). Th e firs t ER w as re garde d as th e only NR th at m e diate s e s troge nic e ffe cts, untila nove ler w as clone d from rat pros tate (14). Th e nove ler, called ER β, is h om ologous to th e firs t ER, now called ER α, particularly in th e D NA binding dom ain (9 7% am ino acid ide ntity) and in th e ligandbinding dom ain (55% am ino acid ide ntity) (Figure 3) (15). Figure 3. Sch e m atic s tructuralorganization of th e h um an e s troge n re ce ptors. Th e ER s s h are s tructuralch aracte ris tics w ith m e m be rs of th e NR s upe rfam ily including five dis tinguis h able dom ains (16). Th e y are nam e d th e A/B, C, D, E and F dom ains, re s pe ctive ly (Figure 3). Th e N-te rm inala/b dom ain is h igh ly variable in s e q ue nce and length be tw e e n diffe re nt NR prote ins and us ually contains a trans activation function th at activate s trans cription of targe t ge ne s. Th e C dom ain, or th e D NA-binding dom ain, is th e m os t cons e rve d re gion be tw e e n diffe re nt m e m be rs of th e fam ily. Th is dom ain is involve d in s pe cific D NA binding and re ce ptor dim e rization. Th e D dom ain w ork s as a flexible h inge be tw e e n th e D NA-binding dom ain and th e E dom ain. Th e E dom ain is re fe rre d to as th e ligand-binding dom ain. It is im portant for ligand binding, re ce ptor dim e rization and 4

11 Es troge n s ignaling in m e tabolic dis e as e ;a functionalge nom ics approach trans criptionalactivation. Th e F-dom ain is not pre s e nt in allnr s and th e functions of th is dom ain are s tillpoorly unde rs tood. Th e tw o ER s h ave s im ilar but not ide nticalligand binding prope rtie s (14). Th e te rm Se lective Es troge n R e ce ptor M odulator (SER M ) w as de ve lope d to provide a m ore ge ne ric de s cription of h ow th e activity of an ER ligand de rive s from its m odulation of ER conform ation. Th e conce pt involve s th e ability to s e lective ly prom ote inte ractions of ER α or ER β, re s pe ctive ly, w ith diffe re nt prote ins s uch as trans criptional coactivators or core pre s s ors (17). Th e pos s ibility to, bas e d on s ubtype or tis s ue s e lective ER activation, s ynth e s ize com pounds w ith im prove d th e rape utic profiles is an are a unde r inte ns e inve s tigation. Tam oxife n, for e xam ple, is an ER agonis t in bone and ute rus, but an ER antagonis tin bre as ttis s ue, and is th e re fore us e d for bre as tcance r tre atm e nt. Ce rtain ch e m icals in th e e nvironm e nt, s uch as naturalplant com pounds and indus trial products, are ligands for ER s. Th e e ndocrine -dis rupting e ffe cts of th e s e ch e m icals and th e ir e ffe cts on h um an h e alth are unde r inte ns ive inve s tigation (18-20). ER s are w ide ly dis tribute d in th e body. ER α is m ainly e xpre s s e d in ute rus, pros tate (s trom a), ovary (th e ca ce lls), te s te s (Le ydig ce lls), e pididym is, bone, bre as t, various re gions of th e brain, live r and w h ite adipos e tis s ue. ER β is e xpre s s e d in for e xam ple colon, pros tate (e pith e lium ), te s tis, ovary (granulos a ce lls), bone m arrow, s alivary gland, vas cular e ndoth e lium and ce rtain re gions of th e brain. Furth e rm ore, in s om e tis s ue s, both ER s are e xpre s s e d albe it in diffe re nt ce lltype s. Th e diffe re nt tis s ue dis tribution for th e tw o ER s m ay accountfor tis s ue s pe cific e ffe cts of e s troge n (12, 21, 22) M e ch anis m s ofe s troge n s ignal ing M os t of th e k now n actions of e s troge n are s o called ge nom ic e ffe cts, w h ich are m e diate d by ER s th rough th e ir inte raction w ith D NA, in diffe re nt propos e d m ode s th at activate or re pre s s th e e xpre s s ion of s pe cific s e ts of ge ne s. Th e re is also e vide nce s ugge s ting th at e s troge n h as non-ge nom ic e ffe cts th at are too rapid to be accounte d for by trans criptional activation or re pre s s ion of targe t ge ne s. Figure 4 is a s um m ary of th e propos e d m ode s for m e ch anis m s of e s troge n s ignaling Ge nom ic e ffe cts ofe s troge n Es troge ns can diffus e acros s th e plas m a and nuclear m e m brane s of ce lls and bind to ER s th at e xis t in a com plex w ith prote ins, including h e at s h ock prote ins, w ith in th e ce llnucleus (23, 24). Binding of ligand activate s ER s, by a m e ch anis m th at involve s dis s ociation of h e at s h ock prote ins and dim e rization of re ce ptor prote ins (25, 26). Th e activate d ER s bind as h om odim e rs or h e te rodim e rs to e s troge n re s pons e e lem e nts (ER Es ). Th e ER E w as firs t ide ntifie d by aligning s e q ue nce s w h ich s h are d s e q ue nce h om ology in th e 5 flank ing 5

12 H ui Gao re gion of th e e s troge n re gulate d vite lloge nin ge ne s A1, A2, B1, B2 from Xe nopus lae vis and ch ick e n (27, 28). Th e ER E s e q ue nce is a 13bp palindrom ic inve rte d re pe at w ith th e cons e ns us s e q ue nce : 5 -GGTCAnnnTGACC-3. Th e ER E w as s h ow n to act on a h e te rologous prom ote r in an orie ntation and dis tance inde pe nde nt m anne r, th e re by fullfilling th e crite ria for an e nh ance r. Th e binding of ER s to ER Es facilitate s th e as s e m bly of bas al trans cription factors into a s table pre -initiation com plex and incre as e s th e trans cription rate for targe tm R NA s ynth e s is (29 ). In addition to ligand m e diate d activation, ER s can also be activate d by e xtrace llular s ignals in th e abs e nce of ligands. Th is m ode lis prim arily de rive d from s tudie s of ER activation by polype ptide grow th factors s uch as EGF (30, 31). Th e biologicals ignificance of th is proce s s is notye tclear (32). In addition to binding to th e ER E, th e activate d ER s can inte ract w ith oth e r D NA-bound trans cription factors to re gulate th e trans cription of ce rtain s e ts of ge ne s, to s om e e xte nt dis tinct from th os e re gulate d via ER Es. In th is m e ch anis m, ER s do not th e m s e lve s bind D NA. AP-1 s ite s and SP-1 s ite s are w e llch aracte rize d m otifs th at could m e diate e s troge n s ignaling via oth e r bound trans cription factors, s uch as JUN/FO S and SP-1 (33-40). Figure 4. Propos e d m e ch anis m s of e s troge n s ignaling. Th e biological e ffe cts of e s troge n are m e diate d th rough at leas t four path w ays. 1, Ligand de pe nde nt via an ER E. 2, Ligand de pe nde nt but ER E inde pe nde nt. 3, Ligand inde pe nde nt, via ER E or ER E inde pe nde nt. 4, Nonge nom ic ER s ignaling Non-ge nom ic e ffe cts ofe s troge n Non-ge nom ic e ffe cts of e s troge n h ave be e n obs e rve d in diffe re nt organs and ce lltype s, 6

13 Es troge n s ignaling in m e tabolic dis e as e ;a functionalge nom ics approach s uch as ute rus and ne urons (41, 42). Non-ge nom ic e ffe cts occur rapidly, w ith in s e conds to m inute s, afte r e s troge n tre atm e nt and cannot be block e d by trans cription and trans lational inh ibitors. Studie s h ave s ugge s te d th at th e s e e ffe cts m ay be th e re s ult of e s troge n activation of M APK and ER K s ignaling (43) or re leas e of intrace llular calcium (44). It h as be e n propos e d th at th e s e e ffe cts are m e diate d by s pe cific re ce ptors on th e plas m a m e m brane of targe t ce lls (43, 45-47). H ow e ve r, s uch m e m brane as s ociate d ER prote ins h ave s o far not be e n conclus ive ly ide ntifie d and th e m olecular m e ch anis m for th e nonge nom ic e ffe cts of e s troge n is s tillunclear (48). Figure 5. Ins ulin s ignaltrans duction path w ays 1.3 ESTRO GEN SIGNALING AND M ETABO LIC DISEASE M e tabol ic dis e as e M e tabolis m re fe rs to m odification of ch e m icalcom pounds in living organis m s and ce lls (49 ), by w h ich th e body ge ne rate s e ne rgy and s ynth e s ize s oth e r e s s e ntialm olecules it ne e ds from fat, carboh ydrate s and prote ins de rive d from our die t. In ge ne ral, m e tabolis m of com pounds include s a s e q ue nce of e nzym atic s te ps, re fe rre d to as m e tabolic path w ays. Th e re is us ually a s ignificant leve lof re dundance in m e tabolic path w ays. Fre q ue ntly, th e re w ill be m ore th an one w ay to obtain a varie ty of m e tabolic inte rm e diate s involving diffe re nt e nzym e s. D is e as e w illonly occur if a criticale nzym e is dis abled or if an e s s e ntial controlm e ch anis m for a m e tabolic path w ay is affe cte d (50). 7

14 H ui Gao Ins ulin is a polype ptide h orm one s e cre te d by th e be ta ce lls of th e pancre as th at re gulate carboh ydrate and fat m e tabolis m, e s pe cially th e conve rs ion of glucos e to glycoge n, w h ich low e rs blood glucos e leve l(figure 5). O ur k now ledge about th e m olecular m e ch anis m of ins ulin s ignaling is s tilllim ite d. Th e initials te p is th e binding of th e ins ulin h orm one to th e m e m brane bound ins ulin re ce ptor on th e ce lls urface. Th e binding of ins ulin to th e ins ulin re ce ptor initiate s a s ignaling cas cade th at ultim ate ly leads to incre as e d glucos e uptak e, re duce d fat bre ak dow n and incre as e d glycoge n s torage in targe t tis s ue s including th e live r, s k e letalm us cle and adipos e tis s ue. M olecular e xe rcis e and w e igh t re duction incre as e ins ulin s e ns itivity w h ilst obe s ity is as s ociate d w ith ins ulin re s is tance. Furth e rm ore, it is be lie ve d th at ge ne tic factors influe nce ins ulin s e ns itivity(51-54) Ins ulin re s is tance Ins ulin re s is tance is th e condition in w h ich norm alam ounts of ins ulin are inade q uate to produce a norm alins ulin re s pons e in fat, m us cle and live r tis s ue s (55). It re s ults in abnorm alitie s in glucos e m e tabolis m, w ith re duce d pe riph e ral dis pos al of glucos e in m us cle and fat and incre as e d h e patic glucos e output in th e fas ting s tate. H igh plas m a leve ls of ins ulin and glucos e due to ins ulin re s is tance ofte n lead to th e m e tabolic s yndrom e and type 2 diabe te s (T2D ). Ins ulin re s is tance h as e m e rge d as a conce pt link ing diabe te s m e llitus and cardiovas cular dis e as e (56-59 ). Th e de ve lopm e nt of th e m e tabolic s yndrom e and its re lation to cardiovas cular dis e as e and diabe te s are s um m arize d in Figure 6. Figure 6. Sum m ary of th e de ve lopm e nt of th e m e tabolic s yndrom e and its re lation to cardiovas cular dis e as e and diabe te s 8

15 Es troge n s ignaling in m e tabolic dis e as e ;a functionalge nom ics approach M e tabolic s yndrom e Th e m e tabolic s yndrom e re fe rs to a group of inte rre late d m e tabolic abnorm alitie s th at include dis turbe d glucos e and ins ulin m e tabolis m, ove rw e igh t and abdom inal fat accum ulation, m ild dys lipide m ia, and h ype rte ns ion (60). Ins ulin re s is tance is be lie ve d to be th e k e y factor for th e path oge ne s is of th e m e tabolic s yndrom e (61). Th e pre valence of th e m e tabolic s yndrom e is rapidly incre as ing w orld w ide. Early diagnos is and tre atm e nt s h ould be a priority to re duce th e ris k for progre s s ion to T2D and cardiovas cular dis e as e (62). Tre atm e nt of m e tabolic s yndrom e involve s im prove m e nt of th e unde rlying ins ulin re s is tance th rough life s tyle m odification s uch as w e igh t re duction and incre as e d ph ys ical activity and in m ore s e ve re cas e s, th e rape utic inte rve ntion Type 2 diabe te s m e llitus T2D is a s e ve re m e tabolic dis e as e ch aracte rize d by pe rs is te nt h ype rglyce m ia and ins ulin re s is tance. It is now cons ide re d to be a m ajor th re at to w orld h e alth. Alth ough th e path oph ys iologicalbas is for T2D is not fully clarifie d, accum ulate d e vide nce s ugge s ts th at m ode rn die ts and s e de ntary life s tyles progre s s ive ly lead to th e accum ulation of s aturate d fatty acids in nonadipos e tis s ue (63), s ucce e de d by de cre as e d ins ulin-s tim ulate d glucos e uptak e in s k e letalm us cle, uns uppre s s e d h e patic glucos e production, and alte re d glucos e s tim ulate d ins ulin re leas e from pancre atic be ta ce lls. T2D is us ually initially tre ate d by atte m pts to ch ange life s tyle. H ow e ve r, th is ofte n ne e ds to be com plem e nte d by antidiabe tic drugs, particularly as th e dis e as e progre s s e s. M any antidiabe tic drugs are available on th e m ark e t and th e s e can be s ubdivide d into s ix groups : ins ulin, s ufonylure as, m e glitinide s, biguanide s, th iazolidine dione s, and alph a-glucos idas e inh ibitors. Th e s e drugs act th rough diffe re nt m e ch anis m s s uch as incre as ing ins ulin s e cre tion (s ulfonylure as, m e tglitinide s ), de cre as ing glucos e re leas e from th e live r (biguanid), incre as ing ins ulin s e ns itivity (th iazoledine dione s ) and de cre as ing th e abs orption of carboh ydrate s from th e inte s tine (alph a glucos idas e inh ibitors ). Exis ting th e rapie s for T2D re m ain inade q uate, and th us th e re is a gre atne e d for de ve lopm e ntof ne w drugs Cardiovas cular dis e as e Cardiovas cular dis e as e is th e leading caus e of de ath in indus trialize d countrie s (64). It re fe rs to a clas s of dis e as e s th at involve th e h e art and/or blood ve s s e ls (arte rie s and ve ins ) s uch as h ype rte ns ion, angina and m yocardialinfarction as w e llas ce re brovas cular dis e as e including s trok e. Th e s e s dis e as e s us ually occur as a re s ult of arte rial dam age. Th e s ym ptom s and tre atm e nts de pe nd on w h ich s e t(or s e ts ) of arte rie s are affe cte d. ER s could pote ntially be targe ts for th e tre atm e nt of T2D and various as pe cts of cardiovas cular dis e as e, particularly as th e activity of th e s e re ce ptors can be re gulate d by s m allm olecular ligands. 9

16 H ui Gao Es troge n and th e controlofgl ucos e m e tabol is m M ore and m ore e vide nce from s tudie s in h um ans and rode nts link e s troge n to th e m ainte nance of glucos e h om e os tas is (5). Pos tm e nopaus alw om e n de ve lop vis ce ralobe s ity, ins ulin re s is tance and are at h igh ris k for T2D. Tre atm e nt of h e alth y pos tm e nopaus al w om e n w ith unoppos e d e s tradiol or conjugate d e q uine e s troge ns h as be e n s h ow n to im prove ins ulin s e ns itivity and to low e r blood glucos e (65-67). Furth e rm ore, a random ize d, double-blind, place bo-controlled trialh as s h ow n th at h orm one re place m e nt th e rapy in pos tm e nopaus alw om e n w ith coronary arte ry dis e as e re s ults in a 35% re duction in th e incide nce of T2D at th e 4-yr follow up (68). Fe m ale rode nts are prote cte d agains t h ype rglyce m ia unles s th e y are ovarie ctom ize d in th e s pontane ous rode nt m ode ls of T2D (69-71). R e ce ntly, s tudie s of k nock out m ice w ith ge ne tic alte ration of e s troge n s ynth e s is (AR K O ) or e s troge n action (ER K O ) h ave provide d additionale vide nce for th e prote ctive role of e s troge n in m aintaining glucos e h om e os tas is and ins ulin s e ns itivity (72, 73). AR K O m ice are de lete d w ith re s pe ct to th e arom atas e ge ne, and th us cannot s ynth e s ize e ndoge nous e s troge ns. In ER K O m ice th e ER α ge ne h as be e n de lete d. Both AR K O and ER K O m ice de ve lop ins ulin re s is tance. Th e live rs from AR K O m ice s h ow e xagge rate d triglyce ride accum ulation and re duce d e xpre s s ion of e nzym e s involve d in pe roxis om aland m itoch ondrialfatty acid oxidation (74). Elucidation of th e s ignaling path w ays involve d in e s troge n action on glucos e m e tabolis m in rode nts m ay be us e fulin ide ntifying ne w drug targe ts to pre ve ntt2d Es troge n and th e cardiovas cul ar s ys te m Cardiovas cular dis e as e is th e leading caus e of de ath in indus trialize d countrie s (64). Th e incide nce of cardiovas cular dis e as e de m ons trate s s ignificant ge nde r-bas e d diffe re nce s, s upporting a role of e s troge n in th e e tiology of th e dis e as e (75). Es troge ns are th ough t to prom ote cardiovas cular prote ction by low e ring plas m a leve ls of low de ns ity lipoprote in ch oles te roland incre as ing plas m a leve ls of h igh de ns ity lipoprote in ch oles te rol, prom oting coronary vas odilatation, im proving glucos e m e tabolis m and de cre as ing s e rum ins ulin leve ls. Additionals tudie s s ugge s t th at dire ct actions of e s troge n on h e art and blood ve s s e ls contribute to th e cardiovas cular prote ctive e ffe cts of e s troge n (76). Th e s tudie s re fe rre d to above s ugge s t th at e s troge n re place m e nt th e rapy (ERT) could re duce th e ris k of coronary h e art dis e as e (CH D ) in pos t-m e nopaus alw om e n. H ow e ve r, th e data from tw o big clinicaltrials, th e H e art and Es troge n/proge s tin R e place m e nt Study (H ER S) and th e W om e n s H e alth Initiative (W H I), indicate th at th e rapy w ith e s troge n alone or com bine d e s troge n and proge s tin re place m e nt th e rapy is not e ffe ctive in pre ve ntion of CH D and m igh t actually incre as e th e ris k of CH D in th e firs t ye ar (77, 78) (79 ). O n th e oth e r h and, s e ve ralauth ors h ave h e avily criticize d in particular th e W om e n s H e alth Initiative s tudy, claim ing th at th e conclus ions draw n w e re not w arrante d, due to th e 10

17 Es troge n s ignaling in m e tabolic dis e as e ;a functionalge nom ics approach flaw e d de s ign of th is s tudy(80-82). Irre s pe ctive of th e im portance of e s troge n in prote cting agains t cardiovas cular dis e as e, th e h orm one w illre m ain a th e rape utic for oth e r indications. Its e ffe cts on th e cardiovas cular s ys te m ne e d to be ch aracte rize d. So far th e re is lim ite d inform ation about e s troge n re gulate d ge ne s in th e h e art. 1.4 STUDYING ESTRO GEN ACTIO N USING A FUNCTIO NAL GENO M ICS APPRO ACH Functionalge nom ics Th e nam e ge nom ics w as coine d by Tom as R ode rick to de s cribe th e s cie ntific dis cipline of m apping, s e q ue ncing, and analyzing ge nom e s (83). In 2001, th e draft of th e com plete h um an ge nom e s e q ue nce w as publis h e d, and one ye ar late r, th e m ous e ge nom e s e q ue nce w as de s cribe d (84, 85). Additionalge nom e s continue to be s e q ue nce d and as s e m bled. Th e com pletion of ge nom e s e q ue ncing for m any organis m s h as s h ifte d th e inte re s t in ge nom ic re s e arch from m apping and s e q ue ncing of ge nom e s to s tudie s aim e d at unde rs tanding th e function of ge nom e s. Th e te rm functional ge nom ics re flects th is trans ition. M ore s pe cifically, it re fe rs to th e de ve lopm e nt and application of ge nom e -w ide e xpe rim e ntal approach e s to as s e s s ge ne function by m ak ing us e of th e inform ation and re age nts provide d by s tructuralge nom ics (m apping and s e q ue ncing of ge nom e s ) (83). Functional ge nom ics focus e s on dynam ic as pe cts s uch as ge ne trans cription, prote in trans lation, and prote in-prote in inte ractions in ce lls, s tudie d in a s ys te m atic m anne r, inve s tigating allge ne s and prote ins at th e s am e tim e. A functionalge nom ics approach is ch aracte rize d by h igh th rough put e xpe rim e ntal m e th odologie s com bine d w ith s tatis tical and com putational analys is of th e data Studying e s troge n s ignal ing in a ge nom e w ide fas h ion In th e pas t 80 ye ars from th e firs t is olation of e s troge ns, num e rous s tudie s h ave addre s s e d th e action of e s troge n and dis cove re d s om e of th e unde rlying s ignaling path w ays th at m e diate e s troge n action (1). Th e ide ntification of th e ER in s h ifte d th e paradigm of s te roid h orm one re s e arch from th e e nzym atic one to a ne w m ode lw h e re s te roids diffus e into ce lls and inte ract w ith s pe cific re ce ptors to e xe rt th e ir biologicalfunction (86). Late r it w as de m ons trate d th at ER prim arily acts as a trans cription factor to re gulate ge ne e xpre s s ion (87). Es pe cially afte r th e cloning of ER s, s cie ntis ts h ave conce ntrate d th e ir e fforts on ide ntifying th e m olecular e ve nts th at m e diate e s troge n s ignaling in diffe re nt tis s ue s, diffe re nt ph ys iological s tate s and diffe re nt dis e as e s. Th e s e e fforts include th e ide ntification of ER targe t ge ne s (88-9 0), ide ntification of ER E (9 1) and ide ntification of cofactors th at are involve d in ER m e diate d trans criptional re gulation ( ). M ore re ce ntly, s tudie s h ave s h ow n th at ER s m ay affe ct th e ch rom atin s tructure of th e ir targe t 11

18 H ui Gao ge ne s (9 5, 9 6). All th e s e e fforts h ave clarifie d num e rous crucial m e ch anis m s and contribute d gre atly to our unde rs tanding of e s troge n s ignaling. H ow e ve r, m any parts of e s troge n s ignaling ne tw ork s s tillre m ain unk now n.. Functionalge nom ics h as provide d ne w pos s ibilitie s to inve s tigate actions of e s troge n in diffe re nt tis s ue s and ce ll type s. Ge ne e xpre s s ion analys is by m icroarray is a h igh th rough put te ch niq ue th at allow s m e as ure m e nts of global ge ne e xpre s s ion in one e xpe rim e nt. It h as be e n us e d to try to dis s e ct th e e s troge n s ignaling path w ay in de tail, to ide ntify e s troge n re s pons e ge ne s afte r ligand activation of ER s ( ) and to ch aracte rize th e tim e cours e of e s troge n re s pons e (100). Th is approach h as lead to th e ide ntification of a num be r of nove le s troge n targe tge ne s. D e s pite s ignificant ach ie ve m e nts in ide ntification of ER re s pons ive ge ne s, little is k now n about th e organization of cis -re gulatory e lem e nts in ER targe t ge ne prom ote rs and inte ractions be tw e e n ER and ch rom atin. R e ce ntly, a h igh th rough put te ch nology com bining ch rom atin im m unopre cipitation (Ch IP) w ith ge nom ic D NA m icroarrays (ch ip) h as be e n us e d to ide ntify loci/s ite s th at are bound by ER α in intact ch rom atin in ce lls (101103). Th is te ch nology is now re fe rre d to as Ch IP-on-ch ip. So far m os t of th e functionalge nom ics s tudie s of e s troge n s ignaling h ave be e n focus e d on s tudie s in bre as t cance r ce llline s. In th is th e s is, I us e functionalge nom ics to inve s tigate th e role of e s troge n in live r and h e art tis s ue and try to corre late th e s e data to ph ys iological obs e rvations. 12

19 Es troge n s ignaling in m e tabolic dis e as e ;a functionalge nom ics approach 2 AIM S O F TH E STUDY Th e ove rallaim of th is th e s is w as to im prove our unde rs tanding of th e m olecular e ve nts involve d in m e diating th e e ffe cts of e s troge n in re lation to glucos e h om e os tas is and h e art function. Th e s pe cific aim s w e re : I. To inve s tigate th e m e ch anis m s of th e e ffe cts of e s troge n on glucos e h om e os tas is us ing ER K O and BER K O m ice (Pape r I). II. To inve s tigate th e e ffe cts of long-te rm e s tradiol adm inis tration on ins ulin s e ns itivity and to e xplore th e m olecular m e ch anis m s th at unde rlie th e antidiabe tic e ffe cts of e s troge n us ing ob/ob m ice as a m ode l(pape r II). III. To de te rm ine ge nom e w ide prom ote r occupancy by ER α in m ous e l ive r and to corre l ate th e s e data to ph ys iologicalobs e rvations (Pape r III). IV. To s tudy th e m olecular m e ch anis m of e s troge n action in th e h e art and to ide ntify nove le s troge n re gulate d ge ne s (Pape r IV). 13

20 H ui Gao 3 M ETH O DO LO GICAL CO NSIDERATIO NS Th e m e th ods us e d in th is th e s is are de s cribe d in th e M ate rials and M e th ods s e ctions in e ach of th e individual pape rs. In th is s e ction I focus on h igh th rough put functional ge nom ic approach e s w h ich are ce ntralto th is th e s is. 3.1 GENE EXPRESSIO N PRO FILING USING M ICRO ARRAYS In ge ne e xpre s s ion m icroarrays, th ous ands of probe s are fixe d to a s urface, and R NA s am ples are labe led w ith fluore s ce nt dye s for h ybridization. Afte r h ybridization, las e r ligh t is us e d to e xcite th e fluore s ce nt e m is s ion, w h ich give s an e s tim ate of th e re lative am ounts of th e diffe re nt trans cripts th at are re pre s e nte d. Th e re are m any m icroarray platform s th at are diffe re nt in array de s ign and dye s e lection (104). Spotte d arrays and th e Affym e trix Ge ne Ch ip arrays are th e m os t popular m icroarray platform s. Affym e trix Ge ne Ch ip is a h igh -de ns ity oligonucleotide m icroarray. It cons is ts of in-s itu ch e m ically s ynth e s ize d probe s of 25 nucleotide s. A probe pair cons is ts of a pe rfe ct m atch (PM ) probe and a m is m atch (M M ) probe. Th e M M probe h as ide nticals e q ue nce to th e PM probe, e xce pt a one bas e diffe re nce at a ce ntralpos ition. Th e M M probe is cons ide re d as an inte rnal controlfor cros s h ybridization for th e PM probe (Figure 7). Afte r conve rting th e m R NA s am ple to Biotin labe led cr NA, one s am ple is h ybridize d to e ach array. Figure 7. Th e Affym e trix Ge ne Ch ip array de s ign. Th e pre s e nce of m R NA is de te cte d by a s e rie s of probe pairs th at diffe r in only one nucleotide, in th is cas e 11. Th e availability of alte rnative platform s for m icroarrays rais e s th e q ue s tion; w h ich is th e be s t platform for one s e xpe rim e nt?th is is an on going de bate. H ow e ve r, re ce nt s tudie s h ave s ugge s te d th at biologicaltre atm e nts h ave a gre ate r im pact on m e as ure d e xpre s s ion 14

21 Es troge n s ignaling in m e tabolic dis e as e ;a functionalge nom ics approach leve ls th an th e ch oice of platform for m ore th an 9 0% of analyze d ge ne s (105). Anoth e r s tudy re porte d th at re producibility for m os t platform s w ith in any laboratory w as typically good, but re producibility be tw e e n platform s and acros s laboratorie s w as ge ne rally poor. R e producibility be tw e e n laboratorie s incre as e d m ark e dly w h e n s tandardize d protocols w e re im plem e nte d for R NA labe ling, h ybridization, m icroarray proce s s ing, data acq uis ition and data norm alization (106). In th is th e s is, w e us e th e Affym e trix Ge ne Ch ip s ys te m to s tudy th e ge ne e xpre s s ion profiles in m ous e live r (pape rs I and II) and h e art (pape r IV). Th e Affym e trix Ge ne Ch ip m urine ge nom e U74A arrays, w ith probe s re pre s e nting ove r trans cripts, w e re us e d for pape r IV and th e Affym e trix Ge ne Ch ip m urine ge nom e 430A arrays, w ith probe s re pre s e nting ove r trans cripts, w e re us e d for pape rs I and II. For th e clos e ly re late d s tudie s in pape rs I and II, itw as im portantto us e th e s am e type of arrays. A typicale xpre s s ion profiling e xpe rim e nt us ing m icroarrays involve s a num be r of s te ps, from initiale xpe rim e ntalde s ign to th e finalanalys is of th e obtaine d data (Figure 8). For th e Affym e trix Ge ne Ch ip s ys te m, th e s te ps involve d in R NA s am ple pre paration, h ybridization and im age analys is are s tandardize d by ins trum e nts, s oftw are, and protocols re leas e d from th e com pany. In th e follow ing, I w illfocus on four k e y s te ps of a ge ne e xpre s s ion profiling e xpe rim e nt us ing th e Affym e trix platform ; e xpe rim e ntal de s ign, pre proce s s ing, infe re nce and biologicalinte rpre tation of th e data. Figure 8. An ove rvie w of th e s tage s involve d in a m icroarray e xpre s s ion e xpe rim e nt. R e produce d from Ph arm acoge nom ics 7 (1), w ith pe rm is s ion of Future M e dicine Ltd. 15

22 H ui Gao Expe rim e ntalde s ign Th e obje ctive of e xpe rim e ntal de s ign is to m ak e th e anal ys is of th e data and th e inte rpre tation of th e re s ul ts as s traigh tforw ard and as pow e rfulas pos s ibl e, give n th e purpos e of th e e xpe rim e nt and th e cons traints of th e e xpe rim e ntalm ate rialand pos s ibl y e conom icalcons traints (107). Th e aim of th e s tudy is th e m os t im portant factor in th e e xpe rim e ntalde s ign. Th e e xpe rim e ntalde s ign and th e data anal ys is m e th od coul d be total l y diffe re nt for diffe re nt aim s. For e xam pl e, if th e aim of an e xpe rim e nt is to ide ntify ge ne s th at are diffe re ntial l y e xpre s s e d be tw e e n tw o s tate s as candidate s for furth e r s tudy, s tatis ticalinfe re nce w il lbe a s uitabl e anal ys is m e th od. O n th e oth e r h and, if th e aim is to buil d ge ne tic ne tw ork s from th e data, cons tituting a s e t of diffe re nt s tate s, s tatis tical cl as s ification w il lbe a m ore s uitabl e anal ys is m e th od. As in al lbiol ogicalre s e arch, biol ogicalre pl icate s are re q uire d in m icroarray s tudie s. But h ow m any biol ogicalre pl icate s are ne ce s s ary and s ufficie nt to addre s s th e aim of a s tudy? Traditional approach e s, us e d for biol ogical and cl inical s tudie s, to anal yze s tatis ticalpow e r, are not w e l ls uite d for m icroarray s tudie s. M icroarray s tudie s te s t m any h ypoth e s e s, us e fal s e dis cove ry rate (FD R ) and ofte n us e cl as s ification te ch niq ue s (108). Se ve ralm e th ods h ave be e n de ve l ope d to e s tim ate th e num be r of re pl icate s and th e s am pl e s ize for cl as s ification in a m icroarray s tudy ( ). For th e m os t com m on type of m icroarray s tudy, w h ich is to ide ntify ge ne s th at are diffe re ntial l y e xpre s s e d be tw e e n tw o s tate s, e vide nce indicate s th at 5 biol ogical re pl icate s pe r group s h oul d be anal yze d (108, 112). W ith fe w e xce ptions, m ore re pl icate s provide gre ate r pow e r. In practice, th e num be r of biol ogicalre pl icate s can be l im ite d by th e h igh cos t of m icroarray anal ys is and by th e difficul ty of obtaining biol ogicalor cl inicals am pl e s. Al lge ne e xpre s s ion s tudie s incl ude d in th is th e s is (pape rs I, II and IV) aim e d at ide ntifying th e e s troge n re s pons e ge ne s in diffe re nt tis s ue s. Be caus e of th e budge t of th e proje ct and th e h igh cos t of th e m icroarray at th e tim e w h e n th e proje ct w as initiate d, w e us e d th re e biol ogicalre pl icate s for th e diffe re nt groups. Th re e biol ogical re pl icate s is th e m inim al num be r for pe rform ing s tatis tical te s ts. In our s tudie s, confirm ations of e xpre s s ion l e ve l s us ing inde pe nde nt te ch nol ogie s w e re incl ude d for al l ge ne s th at w e re im portant to ge ne rate a h ypoth e s is and w h ich form e d th e bas is for furth e r s tudie s. R e al -tim e PCR and W e s te rn bl ot anal ys is are us ual l y us e d for confirm ing th e diffe re nt e xpre s s ion l e ve l s for s e l e cte d ge ne s at th e m R NA and prote in l e ve l s, re s pe ctive l y. If th e s am pl e s and/or th e budge t to purs ue m icroarray anal ys is are l im ite d, pool ing m R NA from biol ogicalre pl icate s is favore d by m any inve s tigators. Pool ing coul d re duce th e biol ogicalcom pone nt of variation, but one outl ie r can yie l d m is l e ading re s ul ts by producing a pois one d pool (108). Fol l ow ing th e de cre as e d cos t for th e com m e rcial m icroarrays, individuals am pl e s h ave be com e m ore w ide l y us e d in m icroarray anal ys is. 16

23 Es troge n s ignaling in m e tabolic dis e as e ;a functionalge nom ics approach Pre proce s s ing Low leve lpre proce s s ing s te ps for th e Affym e trix Ge ne Ch ip as s ay cove r proce s s e s from s canne d im age, to obtaining re liable e s tim ate s for re lative m R NA leve ls of alltrans cripts in allof th e s am ples. Th e follow ing dis cus s ion w illfocus on th re e im portant s te ps ;q uality control,norm alization and e xpre s s ion s um m ary. Th e aim of th e q uality controlis to e valuate th e q uality for any particular array and s cre e n out incom parable arrays. Se ve ralq uality controlm e th ods h ave be e n propos e d (113), but no s pe cific m e th od for q uality controlh as e volve d as a cons e ns us m e th od (108). In our s tudie s, th e s canne d im age files w e re e xam ine d for s cratch e s or une ve nly dis tribute d fluore s ce nt s ignals be fore furth e r analys is. Th e ave rage ge ne inte ns ity s h ould be e ve n for allth e arrays in one s tudy. Norm alization is applie d to com pe ns ate for s ys te m atic te ch nicaldiffe re nce s be tw e e n ch ips by re m oving variations caus e d by e xpe rim e ntalproce dure s. Th e re are m any approach e s for norm alizing th e Affym e trix m icroarray data. M os t of th e s e approach e s us e th e as s um ption th at th e e xpre s s ion leve ls of th e m ajority of th e trans cripts on th e arrays do not ch ange be tw e e n th e diffe re nt s am ples th at are be ing s tudie d. Th e s im ples t approach for norm alizing Affym e trix data is to re -s cale e ach ch ip in an e xpe rim e nt to e q ualize th e ave rage (or total) s ignalinte ns ity acros s allch ips. M AS 5.0, th e Affym e trix analys is s oftw are, us e s th is approach to re -s cale e ach ch ip in an e xpe rim e nt to m ak e its ave rage inte ns ity e q ualto an arbitrarily de fine d targe t inte ns ity. In addition to th e globalinte ns ity norm alization de s cribe d above, th e re are a num be r of m ore s oph is ticate d alte rnative approach e s for norm alizing Affym e trix data. Th e s e m e th ods include invariant-s e t norm alization (114), low e s s norm alization and q uartile norm alization (115). Com paris on of diffe re nt norm alization m e th ods indicate th at com plex m e th ods do not ne ce s s arily pe rform be tte r th an th e s im pler once (116). Be caus e of th e de s ign of th e Affym e trix arrays, w h ich us e s m ultiple probe pairs to re pre s e nt e ach ge ne on th e array (Figure 7), a s ignificant e ffort h as be e n focus e d on de ve loping appropriate algorith m s for cons tructing s ingle e xpre s s ion e s tim ate s bas e d on m ultiple-probe h ybridization data. To date, s e ve ralm e th ods h ave be e n publis h e d, w h ich aim to com bine th e diffe re nt inte ns itie s from various probe s corre s ponding to one trans cript, into a s ingle robus t e s tim ate of trans cript abundance. Th e s e m e th ods include Tuk e y-biw e igh t in M AS 5.0, m e dian polis h in R M A and th e e xpre s s ion inde x in dch ip. O f th e pre s e nt s oftw are ve rs ions, R M A could h ave ce rtain advantage s ove r oth e r m e th ods (116). M AS 5.0 from Affym e trix, w as us e d for norm alization and e xpre s s ion s um m arie s for th e s tudie s pre s e nte d in th is th e s is (pape rs I, II and IV). 17

24 H ui Gao Ide ntifying diffe re ntial l y e xpre s s e d ge ne s Th e ge ne e xpre s s ion profiling s tudie s include d in th is th e s is aim to ide ntify e s troge n re s pons e ge ne s in diffe re nt tis s ue s. Infe re ntials tatis tics, s uch as s tatis ticals ignificance te s ts, rath e r th an clas s ification analys is, is th e corre ct analys is s trate gy for th is purpos e. M icroarray data are ve ry diffe re nt form traditional biological data in th at a typical m icroarray e xpe rim e nt involve s a large num be r of h ypoth e s e s and a lim ite d num be r of re plicate s. Th e re fore, a h igh false -pos itive rate is a com m on problem in ide ntifying diffe re ntially e xpre s s e d ge ne s (116). To date, num e rous s tatis ticalm e th ods h ave be e n de ve lope d to incre as e th e pow e r of th e te s ts and to prope rly h andle th e m ultiple te s ting is s ue in m icroarray data analys is (116). D uring th e tim e w h e n I h ave pe rform e d th e w ork include d in th is th e s is, m icroarray data analys is h as e volve d from an initials im plis tic approach into s oph is ticate d bios tatis tics. In our s tudie s, w e ch os e th e analyticalapproach bas e d on th e availability at th e tim e w h e n pe rform ing th e s tudie s and th e s ufficie ncy of th e approach e s to ans w e r our re s e arch q ue s tions. In pape rs I and IV, w e us e d th e concordance call(117) to ide ntify th e diffe re ntially e xpre s s e d ge ne s. Firs t W ilcoxon s igne d-rank te s ts w e re pe rform e d for e ach ge ne on th e m icroarray us ing M AS5.0 for any com bination of tw o s am ples from diffe re nt e xpe rim e ntalgroups. Th e te s t w illas s ign a ch ange call to th e ge ne w h e n de te cting a s ignificant diffe re nce in e xpre s s ion leve ls be tw e e n s am ples. Subs e q ue ntly, a concordance callw as calculate d bas e d on th e num be r of s ignificant pairw is e com paris ons, of e ach individual w ith all individuals in th e oth e r group (9 com paris ons in totalfor th re e individuals in e ach group). In our s tudie s, ge ne s re gulate d in th e s am e dire ction (e ith e r incre as e d or de cre as e d) in at leas t five of th e nine com paris ons w e re s e lecte d as re gulate d ge ne s. Be caus e th e nine com paris ons are not s tatis tically inde pe nde nt, th is m e th od doe s not re pre s e nt a true s tatis ticalte s t. H ow e ve r, w e obs e rve d th at th is approach h ad a low e r false -pos itive e rror th an th e com m on t-te s t th at w as w ide ly us e d at th at tim e. Th is conclus ion is bas e d on confirm ation of m icroarray data us ing re altim e PCR. Furth e rm ore, w e us ually ge t a good concordance be tw e e n th e diffe re ntially e xpre s s e d ge ne s ide ntifie d by our approach and th e non param e tric te s t, s uch as M ann W h itne y te s t. In pape r II, w e us e a m ore s oph is ticate d s tatis ticalapproach to as s e s s th e diffe re nce in ge ne e xpre s s ion by fitting a probe leve lm ode lto th e data, w h ich is s im ilar to w h at h as be e n de s cribe d in th e re fe re nce (118). Th is approach h as be e n s h ow n to be particularly w e lls uite d for s m alls am ple s ize Biol ogicalinte rpre tation Afte r ide ntifying diffe re ntially e xpre s s e d ge ne s, th e ne xt ph as e of m icroarray data analys is re q uire s th at th e diffe re ntially e xpre s s e d ge ne s are as s ociate d w ith biologicalproce s s e s. R e ading th e re levant biom e dicallite rature s could be a s tarting point, but th is is clearly a daunting tas k cons ide ring th at th e re are curre ntly m illions of re s e arch articles in PubM e d. In th e las t fe w ye ars, te xt-m ining and inform ation-e xtraction approach e s h ave be e n de ve lope d to facilitate th is approach by obtaining re levant inform ation us ing autom atic 18