sbm 2: synbio of protein networks

ppi: common motives

Gpa1

Ste 2

Fus3

Ste 5

Ste20

Ste50

Cdc24

Cdc42

Ste 4

Ste 18

Ste 7

Ste 11

GPCRcomplex

MAPKcomplex

GTP-asecomplex

kinases in red

BMN a real example: the pheromone pathway

Jan 2010 Science - W. Lim et al

changes in protein functions have been associ-ated with domain recombination (5). Second,mutations leading to the fusion of protein-codinggenes may lead to the improper activation ofsignaling networks that result in oncogenic trans-formations (6, 7). Third, fusions of diverse reg-ulatory and catalytic domains can yield syntheticproteins with non-natural input/output relation-ships, both in vitro (8) and in vivo (9–11).

To investigate whether recombination ofsignaling protein domains provides a route forevolutionary innovation, analogous to the swappingof cis-regulatory elements and coding sequences intranscriptional circuits (12–14), we have system-atically determined the effects of domain re-combination on the behavior of a well-understoodsignaling network, the yeast mating pathway(Fig. 1A), and compared it to the effects broughtabout by gene or domain duplication. We used

the domains of 11 proteins belonging to themating pathway to construct a library of 66 re-combinant proteins (Fig. 1B). Specifically, allnative proteins composed of at least two do-mains were split in a manner that separated reg-ulatory and catalytic domains. The split pointswere chosen to ensure that domains were left in-tact and therefore are located within interdomainconnecting regions. We then created a library ofchimeric proteins that includes all possible re-combinations of N-terminal and C-terminal blocksto systematically map the resulting phenotypiceffects (Fig. 1C and fig. S1). Each protein wastransformed into a yeast strain that retained theendogenous copies of the 11 mating pathwaygenes, such that an additional protein (with al-tered domain combination) was added to theexisting network. To distinguish the effects ofdomain recombination from those of gene ordomain duplication, we created three additionalsets of strains (Fig. 1D): In the first one, each ofthe 11 genes analyzed was duplicated; in thesecond one, each of the N- or C-terminal blockswas duplicated; and in the third one, each pos-sible pair of N- and C-terminal blocks were du-plicated and coexpressed (all 66 combinationslacking domain recombination). To prevent any

bias that might be related to differential transcrip-tional control, we expressed all constructs at lowabundance using a 250 base pair segment of theconstitutive cycI promoter.

As a metric for how each additional proteinaltered signaling behavior, we measured thedynamics of mating pathway activation by flowcytometry. A green fluorescent protein (GFP)reporter was controlled by a mating-responsivepromoter from the fus1 gene (15) in an a-type,Dfar1 strain [to prevent cell cycle arrest and theformation of mating projections that could affectflow cytometry measurements (16)]. We mea-sured the intensity of GFP fluorescence beforeand after activation of the mating pathway witha-factor and used those values to calculate thebaseline and slope of activation (Fig. 2A). Thenormalized baseline and slope values for eachvariant in our libraries (relative to wild type)were plotted on a “morphospace” diagram. Geneand domain duplications had little effect on thedynamics of pathway activation (Fig. 2, B and C).Only three domain duplication variants showedchanges, slightly inhibiting pathway activation(variants with lower slopes in Fig. 2C), perhapsby acting as dominant negative fragments. Incontrast, recombination of domains resulted in

1Department of Cellular and Molecular Pharmacology,University of California, San Francisco, 600 16th Street, SanFrancisco, CA 94158, USA. 2Howard Hughes Medical Institute,University of California, San Francisco, CA, USA. 3Center forTheoretical Biology, Peking University, Beijing 100871, China.

*Present address: Illumina, Inc., San Diego, CA 92121, USA.†To whom correspondence should be addressed. E-mail:lim@cmp.ucsf.edu

Yeast Mating Pathway Pathway ComponentDomain Architecture

05etS

Ste2

Gpa Ste18 Cdc24

Ste11

alpha-factor

Ste7

Fus3

Downstream Effects

5etS

Ste4Cdc42

Ste20

GpaI

Ste4

Ste18

Cdc24

Cdc42

Ste20

Ste50

Ste11

Ste5

Ste7

Fus3

CH DH PH PB1

PBD Kinase

SAM

RINGPM PH VWA

SAM RA

Ga

Gb

Gg

GTPase

Kinase

Kinase

Kinase

N-t

Domain Recombination Library

targeting/regulatory domain catalytic domain recombination junction

B CA

GpaI

Ste4

Ste18

Cdc24

Cdc42

Ste20

Ste50

Ste11

Ste5

Ste7

Fus3

CH DH PH PB1

PBD Kinase

SAM

RINGPM PH VWA

SAM RA

Ga

Gb

Gg

GTPase

Kinase

Kinase

Kinase

N-t

MAPKKKKMAPKKK

MAPKK

MAPK

(66 variants)

D

full gene duplication

domain duplication

domain recombination

domain co-expression

wt gene(two-domains)

Types of Genetic Changes Analyzed in this Study

Fig. 1. Design of the recombination library ofprotein domains belonging to the yeast matingpathway. (A) The yeast mating pathway is activatedby binding of the mating pheromone (a-factor) to themembrane receptor Ste2 in “a” cells (or a-factor toSte3 in “a” cells), which causes the dissociation of theG protein alpha subunit (GpaI) from the G beta (Ste4)and gamma (Ste18) complex (20, 25). The scaffoldprotein Ste5 is then recruited to the membrane-localized Ste4, bringing along the MAPKKK Ste11,MAPKK Ste7, and MAPK Fus3. In addition, Ste11 in-teracts with the bridging protein Ste50, which by binding to the smallguanosine triphosphatase (GTPase) Cdc42, positions Ste11 near its upstreamactivator, the PAK kinase Ste20 (26). Activated Ste11 phosphorylates Ste7,which in turn phosphorylates Fus3. The activated MAPK translocates to thenucleus, where it phosphorylates a number of transcription factors, leadingto changes in gene transcription, cell cycle progression, and cell morphologyand culminating in the fusion between “a” and “a” cells. (B) Domain

architecture of the yeast mating signaling pathway components. Regulatorydomains are shown in green; catalytic domains are shown in orange. Fullyannotated domain maps are given in fig. S8. (C) Domain recombinationlibrary. Recombination junctions are depicted as white circles; all possiblerecombinations are shown as red connecting lines. (D) Possible evolutionaryevents analyzed in this work. Gene duplication, domain duplication, domainrecombination, and coexpression of two duplicated domains.

www.sciencemag.org SCIENCE VOL 328 16 APRIL 2010 369

REPORTS

on

April

18,

201

0 w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

modularity 1 - domain split

changes in protein functions have been associ-ated with domain recombination (5). Second,mutations leading to the fusion of protein-codinggenes may lead to the improper activation ofsignaling networks that result in oncogenic trans-formations (6, 7). Third, fusions of diverse reg-ulatory and catalytic domains can yield syntheticproteins with non-natural input/output relation-ships, both in vitro (8) and in vivo (9–11).

To investigate whether recombination ofsignaling protein domains provides a route forevolutionary innovation, analogous to the swappingof cis-regulatory elements and coding sequences intranscriptional circuits (12–14), we have system-atically determined the effects of domain re-combination on the behavior of a well-understoodsignaling network, the yeast mating pathway(Fig. 1A), and compared it to the effects broughtabout by gene or domain duplication. We used

the domains of 11 proteins belonging to themating pathway to construct a library of 66 re-combinant proteins (Fig. 1B). Specifically, allnative proteins composed of at least two do-mains were split in a manner that separated reg-ulatory and catalytic domains. The split pointswere chosen to ensure that domains were left in-tact and therefore are located within interdomainconnecting regions. We then created a library ofchimeric proteins that includes all possible re-combinations of N-terminal and C-terminal blocksto systematically map the resulting phenotypiceffects (Fig. 1C and fig. S1). Each protein wastransformed into a yeast strain that retained theendogenous copies of the 11 mating pathwaygenes, such that an additional protein (with al-tered domain combination) was added to theexisting network. To distinguish the effects ofdomain recombination from those of gene ordomain duplication, we created three additionalsets of strains (Fig. 1D): In the first one, each ofthe 11 genes analyzed was duplicated; in thesecond one, each of the N- or C-terminal blockswas duplicated; and in the third one, each pos-sible pair of N- and C-terminal blocks were du-plicated and coexpressed (all 66 combinationslacking domain recombination). To prevent any

bias that might be related to differential transcrip-tional control, we expressed all constructs at lowabundance using a 250 base pair segment of theconstitutive cycI promoter.

As a metric for how each additional proteinaltered signaling behavior, we measured thedynamics of mating pathway activation by flowcytometry. A green fluorescent protein (GFP)reporter was controlled by a mating-responsivepromoter from the fus1 gene (15) in an a-type,Dfar1 strain [to prevent cell cycle arrest and theformation of mating projections that could affectflow cytometry measurements (16)]. We mea-sured the intensity of GFP fluorescence beforeand after activation of the mating pathway witha-factor and used those values to calculate thebaseline and slope of activation (Fig. 2A). Thenormalized baseline and slope values for eachvariant in our libraries (relative to wild type)were plotted on a “morphospace” diagram. Geneand domain duplications had little effect on thedynamics of pathway activation (Fig. 2, B and C).Only three domain duplication variants showedchanges, slightly inhibiting pathway activation(variants with lower slopes in Fig. 2C), perhapsby acting as dominant negative fragments. Incontrast, recombination of domains resulted in

1Department of Cellular and Molecular Pharmacology,University of California, San Francisco, 600 16th Street, SanFrancisco, CA 94158, USA. 2Howard Hughes Medical Institute,University of California, San Francisco, CA, USA. 3Center forTheoretical Biology, Peking University, Beijing 100871, China.

*Present address: Illumina, Inc., San Diego, CA 92121, USA.†To whom correspondence should be addressed. E-mail:lim@cmp.ucsf.edu

Yeast Mating Pathway Pathway ComponentDomain Architecture

05etS

Ste2

Gpa Ste18 Cdc24

Ste11

alpha-factor

Ste7

Fus3

Downstream Effects

5etS

Ste4Cdc42

Ste20

GpaI

Ste4

Ste18

Cdc24

Cdc42

Ste20

Ste50

Ste11

Ste5

Ste7

Fus3

CH DH PH PB1

PBD Kinase

SAM

RINGPM PH VWA

SAM RA

Ga

Gb

Gg

GTPase

Kinase

Kinase

Kinase

N-t

Domain Recombination Library

targeting/regulatory domain catalytic domain recombination junction

B CA

GpaI

Ste4

Ste18

Cdc24

Cdc42

Ste20

Ste50

Ste11

Ste5

Ste7

Fus3

CH DH PH PB1

PBD Kinase

SAM

RINGPM PH VWA

SAM RA

Ga

Gb

Gg

GTPase

Kinase

Kinase

Kinase

N-t

MAPKKKKMAPKKK

MAPKK

MAPK

(66 variants)

D

full gene duplication

domain duplication

domain recombination

domain co-expression

wt gene(two-domains)

Types of Genetic Changes Analyzed in this Study

Fig. 1. Design of the recombination library ofprotein domains belonging to the yeast matingpathway. (A) The yeast mating pathway is activatedby binding of the mating pheromone (a-factor) to themembrane receptor Ste2 in “a” cells (or a-factor toSte3 in “a” cells), which causes the dissociation of theG protein alpha subunit (GpaI) from the G beta (Ste4)and gamma (Ste18) complex (20, 25). The scaffoldprotein Ste5 is then recruited to the membrane-localized Ste4, bringing along the MAPKKK Ste11,MAPKK Ste7, and MAPK Fus3. In addition, Ste11 in-teracts with the bridging protein Ste50, which by binding to the smallguanosine triphosphatase (GTPase) Cdc42, positions Ste11 near its upstreamactivator, the PAK kinase Ste20 (26). Activated Ste11 phosphorylates Ste7,which in turn phosphorylates Fus3. The activated MAPK translocates to thenucleus, where it phosphorylates a number of transcription factors, leadingto changes in gene transcription, cell cycle progression, and cell morphologyand culminating in the fusion between “a” and “a” cells. (B) Domain

architecture of the yeast mating signaling pathway components. Regulatorydomains are shown in green; catalytic domains are shown in orange. Fullyannotated domain maps are given in fig. S8. (C) Domain recombinationlibrary. Recombination junctions are depicted as white circles; all possiblerecombinations are shown as red connecting lines. (D) Possible evolutionaryevents analyzed in this work. Gene duplication, domain duplication, domainrecombination, and coexpression of two duplicated domains.

www.sciencemag.org SCIENCE VOL 328 16 APRIL 2010 369

REPORTS

on

April

18,

201

0 w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

modularity 2 - domain recombination

66 synthetics?

changes in protein functions have been associ-ated with domain recombination (5). Second,mutations leading to the fusion of protein-codinggenes may lead to the improper activation ofsignaling networks that result in oncogenic trans-formations (6, 7). Third, fusions of diverse reg-ulatory and catalytic domains can yield syntheticproteins with non-natural input/output relation-ships, both in vitro (8) and in vivo (9–11).

To investigate whether recombination ofsignaling protein domains provides a route forevolutionary innovation, analogous to the swappingof cis-regulatory elements and coding sequences intranscriptional circuits (12–14), we have system-atically determined the effects of domain re-combination on the behavior of a well-understoodsignaling network, the yeast mating pathway(Fig. 1A), and compared it to the effects broughtabout by gene or domain duplication. We used

the domains of 11 proteins belonging to themating pathway to construct a library of 66 re-combinant proteins (Fig. 1B). Specifically, allnative proteins composed of at least two do-mains were split in a manner that separated reg-ulatory and catalytic domains. The split pointswere chosen to ensure that domains were left in-tact and therefore are located within interdomainconnecting regions. We then created a library ofchimeric proteins that includes all possible re-combinations of N-terminal and C-terminal blocksto systematically map the resulting phenotypiceffects (Fig. 1C and fig. S1). Each protein wastransformed into a yeast strain that retained theendogenous copies of the 11 mating pathwaygenes, such that an additional protein (with al-tered domain combination) was added to theexisting network. To distinguish the effects ofdomain recombination from those of gene ordomain duplication, we created three additionalsets of strains (Fig. 1D): In the first one, each ofthe 11 genes analyzed was duplicated; in thesecond one, each of the N- or C-terminal blockswas duplicated; and in the third one, each pos-sible pair of N- and C-terminal blocks were du-plicated and coexpressed (all 66 combinationslacking domain recombination). To prevent any

bias that might be related to differential transcrip-tional control, we expressed all constructs at lowabundance using a 250 base pair segment of theconstitutive cycI promoter.

As a metric for how each additional proteinaltered signaling behavior, we measured thedynamics of mating pathway activation by flowcytometry. A green fluorescent protein (GFP)reporter was controlled by a mating-responsivepromoter from the fus1 gene (15) in an a-type,Dfar1 strain [to prevent cell cycle arrest and theformation of mating projections that could affectflow cytometry measurements (16)]. We mea-sured the intensity of GFP fluorescence beforeand after activation of the mating pathway witha-factor and used those values to calculate thebaseline and slope of activation (Fig. 2A). Thenormalized baseline and slope values for eachvariant in our libraries (relative to wild type)were plotted on a “morphospace” diagram. Geneand domain duplications had little effect on thedynamics of pathway activation (Fig. 2, B and C).Only three domain duplication variants showedchanges, slightly inhibiting pathway activation(variants with lower slopes in Fig. 2C), perhapsby acting as dominant negative fragments. Incontrast, recombination of domains resulted in

1Department of Cellular and Molecular Pharmacology,University of California, San Francisco, 600 16th Street, SanFrancisco, CA 94158, USA. 2Howard Hughes Medical Institute,University of California, San Francisco, CA, USA. 3Center forTheoretical Biology, Peking University, Beijing 100871, China.

*Present address: Illumina, Inc., San Diego, CA 92121, USA.†To whom correspondence should be addressed. E-mail:lim@cmp.ucsf.edu

Yeast Mating Pathway Pathway ComponentDomain Architecture

05etS

Ste2

Gpa Ste18 Cdc24

Ste11

alpha-factor

Ste7

Fus3

Downstream Effects

5etS

Ste4Cdc42

Ste20

GpaI

Ste4

Ste18

Cdc24

Cdc42

Ste20

Ste50

Ste11

Ste5

Ste7

Fus3

CH DH PH PB1

PBD Kinase

SAM

RINGPM PH VWA

SAM RA

Ga

Gb

Gg

GTPase

Kinase

Kinase

Kinase

N-t

Domain Recombination Library

targeting/regulatory domain catalytic domain recombination junction

B CA

GpaI

Ste4

Ste18

Cdc24

Cdc42

Ste20

Ste50

Ste11

Ste5

Ste7

Fus3

CH DH PH PB1

PBD Kinase

SAM

RINGPM PH VWA

SAM RA

Ga

Gb

Gg

GTPase

Kinase

Kinase

Kinase

N-t

MAPKKKKMAPKKK

MAPKK

MAPK

(66 variants)

D

full gene duplication

domain duplication

domain recombination

domain co-expression

wt gene(two-domains)

Types of Genetic Changes Analyzed in this Study

Fig. 1. Design of the recombination library ofprotein domains belonging to the yeast matingpathway. (A) The yeast mating pathway is activatedby binding of the mating pheromone (a-factor) to themembrane receptor Ste2 in “a” cells (or a-factor toSte3 in “a” cells), which causes the dissociation of theG protein alpha subunit (GpaI) from the G beta (Ste4)and gamma (Ste18) complex (20, 25). The scaffoldprotein Ste5 is then recruited to the membrane-localized Ste4, bringing along the MAPKKK Ste11,MAPKK Ste7, and MAPK Fus3. In addition, Ste11 in-teracts with the bridging protein Ste50, which by binding to the smallguanosine triphosphatase (GTPase) Cdc42, positions Ste11 near its upstreamactivator, the PAK kinase Ste20 (26). Activated Ste11 phosphorylates Ste7,which in turn phosphorylates Fus3. The activated MAPK translocates to thenucleus, where it phosphorylates a number of transcription factors, leadingto changes in gene transcription, cell cycle progression, and cell morphologyand culminating in the fusion between “a” and “a” cells. (B) Domain

architecture of the yeast mating signaling pathway components. Regulatorydomains are shown in green; catalytic domains are shown in orange. Fullyannotated domain maps are given in fig. S8. (C) Domain recombinationlibrary. Recombination junctions are depicted as white circles; all possiblerecombinations are shown as red connecting lines. (D) Possible evolutionaryevents analyzed in this work. Gene duplication, domain duplication, domainrecombination, and coexpression of two duplicated domains.

www.sciencemag.org SCIENCE VOL 328 16 APRIL 2010 369

REPORTS

on

April

18,

201

0 w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

a wide range of altered dynamic behaviors, withvariants that either prevented pathway activationor led to stronger activation of the mating path-way (Fig. 2D). These altered signaling behav-iors appear to depend on domain recombination,because coexpression of all analogous pairs ofunlinked N- and C-terminal domain blocks hadlimited effects on pathway activation (Fig. 2E).At least for the genes and signaling pathway

analyzed here, gene or domain duplication alonemay contribute little to the immediate diversi-fication of signaling phenotypes; changes inpathway behaviors probably require sequencedivergence of the duplicates [e.g., by neofunction-alization or by differential transcriptional regu-lation of subfunctionalized duplicates (17, 18);see fig. S2]. In contrast, shuffling of domains pro-vides a more direct path to functional divergence

(19), resulting in readily available alterations insignaling behaviors.

Beyond changes in gene expression, acti-vation of the mating pathway leads to a co-ordinated response that arrests cell cycle, alterscell morphology, and ultimately results in thefusion of mating partners (20). To determinewhether changes in reporter gene expression dy-namics caused by domain recombination weremirrored by changes in overall pathway out-come, we measured the efficiency with which“a” strains, expressing domain recombination var-iants, mated with wild-type “a” cells. We focusedon the 10 recombination variants with dynamicbehaviors most different from wild type and fromthe corresponding coexpressed N- and C-domainpair (figs. S3 and S4) and measured the per-centage of “a” cells that successfully matedwhen coincubated with “a” cells (21). Yeaststrains expressing domain recombination var-iants with slopes of pathway activation greaterthan that of wild type mated more efficientlythan did wild-type yeast (Fig. 3A and table S1).The same was true for one variant with highbaseline of pathway activation but slightly lowerslope (Ste4[N]-Ste5[C]). In contrast, yeast strainsexpressing variants with activation slopes lowerthan that of wild type mated more poorly. Theobserved changes in mating efficiency alsoappeared to depend on domain recombination,because there were marked differences betweenthe mating efficiencies of corresponding recom-bination and coexpression variants (Fig. 3B).Thus, domain recombination can alter complexpathway outputs, such as the biochemical andmorphological changes needed for mating. Atleast under laboratory conditions, recombinationof protein domains can lead to strains that matemore efficiently than wild type, although furtherwork is needed to determine whether the changesin mating efficiency we observed could confer aselective advantage.

Activation of the mating pathway responsealters the regulation of the cell cycle (16). Inaddition, the mating pathway shares severalproteins with other signaling pathways, such asthe high osmolarity pathway. Thus, domain

Fig. 3. Domain re-combination can leadto strains that matemore efficiently thanwild type. (A) Matingefficiencies were mea-sured for recombina-tion variants with slopeand baseline valuesthat were substantiallydifferent from wild type(>1 SD) and also differ-ent from the slope andbaseline values of the corresponding coexpressed N and C pair (figs. S3and S4). Mating efficiencies of wild type and recombination variants aredepicted as circles, with areas representing relative mating efficiencies. (B)Comparison of domain recombination to coexpression of the correspondingdomain pairs (wild-type values are set to 1). Ste50 SAM domain inter-

acts with the Ste11 (MAPKKK) N-terminal SAM domain, facilitating theinteraction of Ste11 with Ste20, its upstream activator. Thus, it is possiblethat, as an isolated domain, Ste50[N] (as well as Ste11[N]) act as dom-inant negatives, competing for the interaction between the wild-typeproteins.

BA

WT

0

1

2

3

0 1 2 3

Ste5 [N]-Ste11 [C]

Ste4 [N]-Ste5 [C]

Ste50 [N]-Ste7 [C]

Ste20 [N]-Fus3 [C] Ste50 [N]-Ste20 [C]

Ste20 [N]-Ste11 [C]

Cdc42 [N]-Ste18 [C]

Ste50 [N]-Cdc24 [C]

Ste50 [N]-Ste11 [C]

Cdc42 [N]-Ste18 [C]

Ste50 [N]-Ste11 [C]

Ste4 [N]-Ste5 [C]

Ste50 [N]-Cdc24 [C]

Ste5 [N]-Ste11 [C]

Ste50 [N]-Ste7 [C]

0 1 2 3 4

Recombination

Co-expression

Slo

pe (r

elat

ive

to W

T)

Baseline (relative to WT) Mating Efficiency (relative to WT)

Circle A

rea=Mating E

fficiency (relative to W

T)

Fig. 2. Recombinationof protein domains resultsin diversification in signal-ing behaviors. (A) Matingpathway activation wasmeasured by flow cytom-etry, using a GFP reporterunder the control of apromoter (from the Fus1gene) that responds topathway activation, in ana-type DFar1 strain. Timecourse measurements ofGFP fluorescence weredone to calculate thebaseline and slope ofpathway activation underconditions of linear path-way response. Baselineand slopes were thennormalized relative towild-type values, and theresulting values wereplotted in pathway mor-phospace. (B) Gene dupli-cations had minor effectson mating pathway re-sponse dynamics, withmost values clusteredaround wild type. (C)Domain duplicationsalso had minor effects on mating pathway response dynamics; duplication of Ste50[N] (Ste50’s SAMdomain), Ste5[N] (which includes Ste5’s RING domain), and Ste11[N] (Ste11’s SAM domain) areexceptions with low slopes and may act as dominant negative. (D) Domain recombination led to a diverseset of novel signaling behaviors. Recombination variants with dynamic behaviors most different from wildtype and from the corresponding coexpressed N- and C-domain pair (fig. S3) are shown in red. (E) Thesebehaviors could not be recapitulated by coexpression of the unlinked corresponding pairs of domains.Fluorescent values were measured in at least two independent experiments, each time in triplicate. Errorbars: mean values T SD.

Aalphafactor

MatingPathway pFus1-GFP

Time (min)

Slope

Baseline

alpha-factoradded

Baseline (rel.)

PossibleBehaviors

ResponseMorphospace

Time

WT

12

3

INPUT OUTPUT

WT1

23

B

0

1

2

3 Whole Gene Duplication (11) Domain Duplication (12)C

D E

0 1 2 3

0

1

2

3Co-Expression (66)

0

1

2

3

0 1 2 3 0 1 2 3

0 1 2 3

Recombination (66)

0

1

2

3

tuptuO ya

whtaP

)ecnecseroulF PF

G(

tuptuO ya

whtaP

).ler( epolS

N C

N C N C

N

Slo

pe (r

elat

ive

to W

T)

Slo

pe (r

elat

ive

to W

T)

Slo

pe (r

elat

ive

to W

T)

Slo

pe (r

elat

ive

to W

T)

Baseline (relative to WT)

Baseline (relative to WT)

Baseline (relative to WT)

Baseline (relative to WT)

16 APRIL 2010 VOL 328 SCIENCE www.sciencemag.org370

REPORTS

on

April

18,

201

0 w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

morphospace

a wide range of altered dynamic behaviors, withvariants that either prevented pathway activationor led to stronger activation of the mating path-way (Fig. 2D). These altered signaling behav-iors appear to depend on domain recombination,because coexpression of all analogous pairs ofunlinked N- and C-terminal domain blocks hadlimited effects on pathway activation (Fig. 2E).At least for the genes and signaling pathway

analyzed here, gene or domain duplication alonemay contribute little to the immediate diversi-fication of signaling phenotypes; changes inpathway behaviors probably require sequencedivergence of the duplicates [e.g., by neofunction-alization or by differential transcriptional regu-lation of subfunctionalized duplicates (17, 18);see fig. S2]. In contrast, shuffling of domains pro-vides a more direct path to functional divergence

(19), resulting in readily available alterations insignaling behaviors.

Beyond changes in gene expression, acti-vation of the mating pathway leads to a co-ordinated response that arrests cell cycle, alterscell morphology, and ultimately results in thefusion of mating partners (20). To determinewhether changes in reporter gene expression dy-namics caused by domain recombination weremirrored by changes in overall pathway out-come, we measured the efficiency with which“a” strains, expressing domain recombination var-iants, mated with wild-type “a” cells. We focusedon the 10 recombination variants with dynamicbehaviors most different from wild type and fromthe corresponding coexpressed N- and C-domainpair (figs. S3 and S4) and measured the per-centage of “a” cells that successfully matedwhen coincubated with “a” cells (21). Yeaststrains expressing domain recombination var-iants with slopes of pathway activation greaterthan that of wild type mated more efficientlythan did wild-type yeast (Fig. 3A and table S1).The same was true for one variant with highbaseline of pathway activation but slightly lowerslope (Ste4[N]-Ste5[C]). In contrast, yeast strainsexpressing variants with activation slopes lowerthan that of wild type mated more poorly. Theobserved changes in mating efficiency alsoappeared to depend on domain recombination,because there were marked differences betweenthe mating efficiencies of corresponding recom-bination and coexpression variants (Fig. 3B).Thus, domain recombination can alter complexpathway outputs, such as the biochemical andmorphological changes needed for mating. Atleast under laboratory conditions, recombinationof protein domains can lead to strains that matemore efficiently than wild type, although furtherwork is needed to determine whether the changesin mating efficiency we observed could confer aselective advantage.

Activation of the mating pathway responsealters the regulation of the cell cycle (16). Inaddition, the mating pathway shares severalproteins with other signaling pathways, such asthe high osmolarity pathway. Thus, domain

Fig. 3. Domain re-combination can leadto strains that matemore efficiently thanwild type. (A) Matingefficiencies were mea-sured for recombina-tion variants with slopeand baseline valuesthat were substantiallydifferent from wild type(>1 SD) and also differ-ent from the slope andbaseline values of the corresponding coexpressed N and C pair (figs. S3and S4). Mating efficiencies of wild type and recombination variants aredepicted as circles, with areas representing relative mating efficiencies. (B)Comparison of domain recombination to coexpression of the correspondingdomain pairs (wild-type values are set to 1). Ste50 SAM domain inter-

acts with the Ste11 (MAPKKK) N-terminal SAM domain, facilitating theinteraction of Ste11 with Ste20, its upstream activator. Thus, it is possiblethat, as an isolated domain, Ste50[N] (as well as Ste11[N]) act as dom-inant negatives, competing for the interaction between the wild-typeproteins.

BA

WT

0

1

2

3

0 1 2 3

Ste5 [N]-Ste11 [C]

Ste4 [N]-Ste5 [C]

Ste50 [N]-Ste7 [C]

Ste20 [N]-Fus3 [C] Ste50 [N]-Ste20 [C]

Ste20 [N]-Ste11 [C]

Cdc42 [N]-Ste18 [C]

Ste50 [N]-Cdc24 [C]

Ste50 [N]-Ste11 [C]

Cdc42 [N]-Ste18 [C]

Ste50 [N]-Ste11 [C]

Ste4 [N]-Ste5 [C]

Ste50 [N]-Cdc24 [C]

Ste5 [N]-Ste11 [C]

Ste50 [N]-Ste7 [C]

0 1 2 3 4

Recombination

Co-expression

Slo

pe (r

elat

ive

to W

T)

Baseline (relative to WT) Mating Efficiency (relative to WT)

Circle A

rea=Mating E

fficiency (relative to W

T)

Fig. 2. Recombinationof protein domains resultsin diversification in signal-ing behaviors. (A) Matingpathway activation wasmeasured by flow cytom-etry, using a GFP reporterunder the control of apromoter (from the Fus1gene) that responds topathway activation, in ana-type DFar1 strain. Timecourse measurements ofGFP fluorescence weredone to calculate thebaseline and slope ofpathway activation underconditions of linear path-way response. Baselineand slopes were thennormalized relative towild-type values, and theresulting values wereplotted in pathway mor-phospace. (B) Gene dupli-cations had minor effectson mating pathway re-sponse dynamics, withmost values clusteredaround wild type. (C)Domain duplicationsalso had minor effects on mating pathway response dynamics; duplication of Ste50[N] (Ste50’s SAMdomain), Ste5[N] (which includes Ste5’s RING domain), and Ste11[N] (Ste11’s SAM domain) areexceptions with low slopes and may act as dominant negative. (D) Domain recombination led to a diverseset of novel signaling behaviors. Recombination variants with dynamic behaviors most different from wildtype and from the corresponding coexpressed N- and C-domain pair (fig. S3) are shown in red. (E) Thesebehaviors could not be recapitulated by coexpression of the unlinked corresponding pairs of domains.Fluorescent values were measured in at least two independent experiments, each time in triplicate. Errorbars: mean values T SD.

Aalphafactor

MatingPathway pFus1-GFP

Time (min)

Slope

Baseline

alpha-factoradded

Baseline (rel.)

PossibleBehaviors

ResponseMorphospace

Time

WT

12

3

INPUT OUTPUT

WT1

23

B

0

1

2

3 Whole Gene Duplication (11) Domain Duplication (12)C

D E

0 1 2 3

0

1

2

3Co-Expression (66)

0

1

2

3

0 1 2 3 0 1 2 3

0 1 2 3

Recombination (66)

0

1

2

3

tuptuO ya

whtaP

)ecnecseroulF PF

G(

tuptuO ya

whtaP

).ler( epolS

N C

N C N C

N

Slo

pe (r

elat

ive

to W

T)

Slo

pe (r

elat

ive

to W

T)

Slo

pe (r

elat

ive

to W

T)

Slo

pe (r

elat

ive

to W

T)

Baseline (relative to WT)

Baseline (relative to WT)

Baseline (relative to WT)

Baseline (relative to WT)

16 APRIL 2010 VOL 328 SCIENCE www.sciencemag.org370

REPORTS

on

April

18,

201

0 w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

results - signalling

a wide range of altered dynamic behaviors, withvariants that either prevented pathway activationor led to stronger activation of the mating path-way (Fig. 2D). These altered signaling behav-iors appear to depend on domain recombination,because coexpression of all analogous pairs ofunlinked N- and C-terminal domain blocks hadlimited effects on pathway activation (Fig. 2E).At least for the genes and signaling pathway

analyzed here, gene or domain duplication alonemay contribute little to the immediate diversi-fication of signaling phenotypes; changes inpathway behaviors probably require sequencedivergence of the duplicates [e.g., by neofunction-alization or by differential transcriptional regu-lation of subfunctionalized duplicates (17, 18);see fig. S2]. In contrast, shuffling of domains pro-vides a more direct path to functional divergence

(19), resulting in readily available alterations insignaling behaviors.

Beyond changes in gene expression, acti-vation of the mating pathway leads to a co-ordinated response that arrests cell cycle, alterscell morphology, and ultimately results in thefusion of mating partners (20). To determinewhether changes in reporter gene expression dy-namics caused by domain recombination weremirrored by changes in overall pathway out-come, we measured the efficiency with which“a” strains, expressing domain recombination var-iants, mated with wild-type “a” cells. We focusedon the 10 recombination variants with dynamicbehaviors most different from wild type and fromthe corresponding coexpressed N- and C-domainpair (figs. S3 and S4) and measured the per-centage of “a” cells that successfully matedwhen coincubated with “a” cells (21). Yeaststrains expressing domain recombination var-iants with slopes of pathway activation greaterthan that of wild type mated more efficientlythan did wild-type yeast (Fig. 3A and table S1).The same was true for one variant with highbaseline of pathway activation but slightly lowerslope (Ste4[N]-Ste5[C]). In contrast, yeast strainsexpressing variants with activation slopes lowerthan that of wild type mated more poorly. Theobserved changes in mating efficiency alsoappeared to depend on domain recombination,because there were marked differences betweenthe mating efficiencies of corresponding recom-bination and coexpression variants (Fig. 3B).Thus, domain recombination can alter complexpathway outputs, such as the biochemical andmorphological changes needed for mating. Atleast under laboratory conditions, recombinationof protein domains can lead to strains that matemore efficiently than wild type, although furtherwork is needed to determine whether the changesin mating efficiency we observed could confer aselective advantage.

Activation of the mating pathway responsealters the regulation of the cell cycle (16). Inaddition, the mating pathway shares severalproteins with other signaling pathways, such asthe high osmolarity pathway. Thus, domain

Fig. 3. Domain re-combination can leadto strains that matemore efficiently thanwild type. (A) Matingefficiencies were mea-sured for recombina-tion variants with slopeand baseline valuesthat were substantiallydifferent from wild type(>1 SD) and also differ-ent from the slope andbaseline values of the corresponding coexpressed N and C pair (figs. S3and S4). Mating efficiencies of wild type and recombination variants aredepicted as circles, with areas representing relative mating efficiencies. (B)Comparison of domain recombination to coexpression of the correspondingdomain pairs (wild-type values are set to 1). Ste50 SAM domain inter-

acts with the Ste11 (MAPKKK) N-terminal SAM domain, facilitating theinteraction of Ste11 with Ste20, its upstream activator. Thus, it is possiblethat, as an isolated domain, Ste50[N] (as well as Ste11[N]) act as dom-inant negatives, competing for the interaction between the wild-typeproteins.

BA

WT

0

1

2

3

0 1 2 3

Ste5 [N]-Ste11 [C]

Ste4 [N]-Ste5 [C]

Ste50 [N]-Ste7 [C]

Ste20 [N]-Fus3 [C] Ste50 [N]-Ste20 [C]

Ste20 [N]-Ste11 [C]

Cdc42 [N]-Ste18 [C]

Ste50 [N]-Cdc24 [C]

Ste50 [N]-Ste11 [C]

Cdc42 [N]-Ste18 [C]

Ste50 [N]-Ste11 [C]

Ste4 [N]-Ste5 [C]

Ste50 [N]-Cdc24 [C]

Ste5 [N]-Ste11 [C]

Ste50 [N]-Ste7 [C]

0 1 2 3 4

Recombination

Co-expression

Slo

pe (r

elat

ive

to W

T)

Baseline (relative to WT) Mating Efficiency (relative to WT)

Circle A

rea=Mating E

fficiency (relative to W

T)

Fig. 2. Recombinationof protein domains resultsin diversification in signal-ing behaviors. (A) Matingpathway activation wasmeasured by flow cytom-etry, using a GFP reporterunder the control of apromoter (from the Fus1gene) that responds topathway activation, in ana-type DFar1 strain. Timecourse measurements ofGFP fluorescence weredone to calculate thebaseline and slope ofpathway activation underconditions of linear path-way response. Baselineand slopes were thennormalized relative towild-type values, and theresulting values wereplotted in pathway mor-phospace. (B) Gene dupli-cations had minor effectson mating pathway re-sponse dynamics, withmost values clusteredaround wild type. (C)Domain duplicationsalso had minor effects on mating pathway response dynamics; duplication of Ste50[N] (Ste50’s SAMdomain), Ste5[N] (which includes Ste5’s RING domain), and Ste11[N] (Ste11’s SAM domain) areexceptions with low slopes and may act as dominant negative. (D) Domain recombination led to a diverseset of novel signaling behaviors. Recombination variants with dynamic behaviors most different from wildtype and from the corresponding coexpressed N- and C-domain pair (fig. S3) are shown in red. (E) Thesebehaviors could not be recapitulated by coexpression of the unlinked corresponding pairs of domains.Fluorescent values were measured in at least two independent experiments, each time in triplicate. Errorbars: mean values T SD.

Aalphafactor

MatingPathway pFus1-GFP

Time (min)

Slope

Baseline

alpha-factoradded

Baseline (rel.)

PossibleBehaviors

ResponseMorphospace

Time

WT

12

3

INPUT OUTPUT

WT1

23

B

0

1

2

3 Whole Gene Duplication (11) Domain Duplication (12)C

D E

0 1 2 3

0

1

2

3Co-Expression (66)

0

1

2

3

0 1 2 3 0 1 2 3

0 1 2 3

Recombination (66)

0

1

2

3tuptu

O yawhta

P)ecnecseroulF

PFG(

tuptuO ya

whtaP

).ler( epolS

N C

N C N C

N

Slo

pe (r

elat

ive

to W

T)

Slo

pe (r

elat

ive

to W

T)

Slo

pe (r

elat

ive

to W

T)

Slo

pe (r

elat

ive

to W

T)

Baseline (relative to WT)

Baseline (relative to WT)

Baseline (relative to WT)

Baseline (relative to WT)

16 APRIL 2010 VOL 328 SCIENCE www.sciencemag.org370

REPORTS

on

April

18,

201

0 w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

results - mating

Gpa1

Ste 2

Fus3

Ste 5

Ste20

Ste50

Cdc24

Cdc42

Ste 4

Ste 18

Ste 7

Ste 11

05:11 = Ste4bs:Ste11kdin green interactionsinherited by 05:11

Ste5N---

Ste11C

the blind watch making game