DEAF-1 regulates immunity gene expressionin DrosophilaDarien E. Reed*, Xinhua M. Huang†, James A. Wohlschlegel†, Michael S. Levine‡§, and Kate Senger§¶

*Helen Diller Family Comprehensive Cancer Center and Cancer Research Institute, University of California, San Francisco, CA 94115-0128; †Departmentof Biological Chemistry, David Geffen School of Medicine, University of California, Los Angeles, CA 90095-1737; ‡Department of Molecular and CellBiology, Division of Genetics, Genomics, and Development, Center for Integrative Genomics, University of California, Berkeley, CA 94720-3200;and ¶Department of Immunology Discovery, Genentech, South San Francisco, CA 94080-4990

Contributed by Michael S. Levine, March 31, 2008 (sent for review February 2, 2008)

Immunity genes are activated in the Drosophila fat body by Rel andGATA transcription factors. Here, we present evidence that anadditional regulatory factor, deformed epidermal autoregulatoryfactor-1 (DEAF-1), also contributes to the immune response and isspecifically important for the induction of two genes encodingantimicrobial peptides, Metchnikowin (Mtk) and Drosomycin (Drs).The systematic mutagenesis of a minimal Mtk 5� enhancer identi-fied a sequence motif essential for both a response to LPS prepa-rations in S2 cells and activation in the larval fat body in responseto bacterial infection. Using affinity chromatography coupled tomultidimensional protein identification technology (MudPIT), weidentified DEAF-1 as a candidate regulator. DEAF-1 activates theexpression of Mtk and Drs promoter-luciferase fusion genes in S2cells. SELEX assays and footprinting data indicate that DEAF-1binds to and activates Mtk and Drs regulatory DNAs via a TTCGGBTmotif. The insertion of this motif into the Diptericin (Dpt) regula-tory region confers DEAF-1 responsiveness to this normally DEAF-1-independent enhancer. The coexpression of DEAF-1 with Dorsal,Dif, and Relish results in the synergistic activation of transcription.We propose that DEAF-1 is a regulator of Drosophila immunity.

Drosomycin � Metchnikowin � MudPIT � transcription

Transcriptional regulation of Drosophila antimicrobial genesdepends on Rel and GATA transcription factors (1–6).

Many immunity genes contain tightly linked Rel- and GATA-binding sites in promoter-proximal regions. GATA sites areimportant for establishing responses in distinct tissues such as thefat body and midgut. Serpent (dGATAb) is thought to regulateantimicrobial gene expression in the fat body (7, 8), whereasdGATAe activates such genes in the midgut in response toingested microbes (9). In contrast, Dorsal, Dif, and Relish, theNF-�B homologues in flies, shuttle between the cytoplasmic andnuclear compartments, acting as ‘‘on/off switches’’ for induction(10–12). Additional factors, such as HOX and POU domainproteins, bind to distal enhancer elements and maintain consti-tutive domains of gene activity (13). A regulatory element (R1)also has been described within the CecA1 enhancer (14), al-though the factor that interacts with this motif is unknown.

Deformed epidermal autoregulatory factor-1 (DEAF-1) is atranscription factor that was originally shown to bind the autoreg-ulatory enhancer of the Deformed (Dfd) Hox gene, which isactivated in embryonic head segments of Drosophila (15). DEAF-1recognizes several TTCG motifs within the portion of the Dfdautoregulatory region termed ‘‘module E.’’ In addition, DEAF-1binds several similar motifs within a Dfd response element (DRE)from the 1.28 gene that enhances maxillary gene expression duringembryogenesis (16). The DEAF-1 binding elements identified inthese studies are reportedly not required for enhancer activityhowever (16, 17).

The 576-aa DEAF-1 protein possesses two conserved do-mains, SAND and MYND. The 113-aa SAND domain (namedfor SP100, AIRE-1, NucP41/75, and DEAF-1) (18) is responsi-ble for DNA binding via a highly conserved KDWK peptidemotif (19, 20). The 32-aa MYND domain (for myeloid, Nervy,

and Deaf-1) contains non-DNA-binding zinc fingers that arethought to mediate protein–protein interactions (15). DEAF-1 ismaternally expressed, and the encoded protein is broadly dis-tributed throughout the early embryo. It exhibits augmentedexpression in the CNS after stage 14 (15). Zygotic mutantsdevelop to pupal stages, but do not eclose, whereas maternalmutants display severe defects in early embryonic patterning(21). Overexpression of DEAF-1 by using a maternal driverinhibits germ-band retraction and causes defects in dorsal closure,whereas overexpression at later stages causes cell death (21).

In vertebrates, the closest relatives of DEAF-1 are nuclearDEAF-1-related factor (NUDR) and Suppressin (SPN) (22, 23).Both factors are expressed in a wide variety of tissue types.NUDR functions to either activate (22) or repress (24) tran-scription depending on its context, and it binds sequencesbearing TTCGGG or TTTCCG motifs (22). SPN does not havea characterized role in transcription. It was originally identifiedas a protein secreted by the bovine pituitary gland that, whenadded to tissue culture media, inhibits splenocyte proliferation(25) and stimulates IFN-�/� production in leukocytes (26).

Previous studies identified a 208-bp proximal enhancer thatregulates the expression of the Mtk gene. This enhancer directshigh levels of transcription in the fat bodies of infected larvae andalso is induced by LPS preparations in S2 cells. These regulatoryactivities depend on a cluster of Rel- and GATA-binding sites.Here, we present evidence that an additional sequence motif(E8) contributes to Mtk activation. Enhancer DNA affinitychromatography assays and proteomic analysis identifiedDEAF-1 as a protein that interacts with the E8 motif. DEAF-1binds to the consensus sequence TTCGGBT, which is containedwithin the E8 region of the Mtk enhancer. Additional DEAF-1consensus motifs are found in the regulatory regions of otherimmunity genes, such as Drosomycin (Drs). Evidence is pre-sented that DEAF-1 works synergistically with Dorsal, Dif, andRelish to induce gene expression in response to LPS. We proposethat DEAF-1 is an essential component of the immune responsein Drosophila.

Results and DiscussionIdentification of a cis-Regulatory Element in the Mtk Enhancer. Toidentify regulatory motifs within the minimal 208-bp Mtk en-hancer, we scrambled the nucleotide sequences of 11 regions(E1–E11) flanking the previously identified Rel- and GATA-binding sites (Fig. 1A). Several of these scrambled elements

Author contributions: D.E.R. and K.S. designed research; D.E.R., X.M.H., J.A.W., and K.S.performed research; X.M.H. and J.A.W. contributed new reagents/analytic tools; D.E.R.,X.M.H., J.A.W., and K.S. analyzed data; and D.E.R., M.S.L., and K.S. wrote the paper.

The authors declare no conflict of interest.

§To whom correspondence may be addressed. E-mail: mlevine@berkeley.edu orsenger.kate@gene.com.

This article contains supporting information online at www.pnas.org/cgi/content/full/0802921105/DCSupplemental.

© 2008 by The National Academy of Sciences of the USA

www.pnas.org�cgi�doi�10.1073�pnas.0802921105 PNAS � June 17, 2008 � vol. 105 � no. 24 � 8351–8356

IMM

UN

OLO

GY

(SEs) were found to alter the activities of Mtk-Luciferase(Mtk-Luc) promoter-reporter fusion genes in transient transfec-tion assays with S2 cells in the presence of LPS (Fig. 1B). Itshould be emphasized that this assay does not measure aresponse to LPS, but rather to Gram-negative peptidoglycansthat commonly occur in commercial preparations of LPS (27,28). Gram-negative peptidoglycans have been shown to signalthrough the Imd and, to a lesser degree, the Toll pathway (27),so this assay probably reflects both types of signaling cascades.Fusion genes bearing scrambled sequences in region 4, 5, or 6 areroughly twice as active as the native enhancer, suggesting adisruption of potential repressor elements. A 50% reduction ininduction is seen for the fusion gene containing scrambledsequences in region 9, suggesting the loss of a weak activatorelement. Most notably, there is a severe 21-fold decrease in theinduced expression of the Mtk-Luc fusion gene containingscrambled sequences in region 8 (SE8). There also is a 3-foldreduction in the constitutive activities of this fusion gene. Thus,region 8 appears to contain an essential activator element.

To test the activities of the E8 sequence in vivo, we examinedtransgenic larvae carrying an Mtk SE8-LacZ transgene. Upon septicinjury with a mixture of Escherichia coli and Micrococcus luteus, thewild-type Mtk-lacZ transgene drives intense lacZ expression in thefat body (Fig. 1C Upper). Mutation of E8 essentially abolishesreporter gene expression in three of four independent lines andallows only a weak response in the fourth (Fig. 1C Lower) (data notshown). The wild-type Mtk-lacZ fusion gene is constitutively activein the posterior proventriculus of most larvae and in the anteriormidgut of �20% of the larvae. Upon ingestion of Erwinia caroto-vora, there is at least a doubling in the number of larvae that exhibitexpression in the anterior midgut (Fig. 1D Upper). In contrast, theMtk SE8-LacZ transgene completely lacks both constitutive andinduced activity throughout the midgut in three of four transgeniclines, with a weak response in the fourth line (Fig. 1D Lower) (datanot shown).

Characterization of DEAF-1-Recognition Sequences. We used a pro-teomics approach to identify proteins that bind E8. The entireMtk regulatory domain (WT and SE8) was biotinylated andcoupled to magnetic Dynal beads (Fig. 2A). An EcoRI restric-tion site was included proximal to the biotin moiety. Dynalbead–DNA complexes were incubated with nuclear extractsfrom S2 cells that had been treated with LPS. The resultingnucleoprotein complexes were washed extensively and elutedwith a brief EcoRI digest. Eluted proteins were then subjectedto Multidimensional Protein Identification Technology (Mud-PIT) analysis (29).

MudPIT identified several candidate proteins that wereuniquely present in the eluate from the native Mtk regulatorysequence and not from the SE8 mutant enhancer sequence. Oneof these, DEAF-1, was particularly interesting because it recog-nizes a sequence motif, TTCG (15, 16), which resembles the E8sequence (TCATTCGGC). This led us to pursue the role ofDEAF-1 in regulating Mtk expression.

To test whether DEAF-1 recognizes Mtk regulatory sequences,gel shift assays were performed with increasing amounts of recom-binant DEAF-1 protein and radiolabeled E8 and SE8 oligonucle-otides (Fig. 2B). The DEAF-1 protein binds to E8 (lanes 1–4), butnot the SE8 scrambled sequence (lanes 5–8). Competition assayswere done by incubating DEAF-1 with the radiolabeled E8 se-quence, followed by the addition of an excess of unlabeled E8 orSE8 oligonucleotides (Fig. 2C). A 10-fold excess of unlabeled E8removes DEAF-1 from the radioactive probe (compare lane 5 withlane 1), whereas the same amount of the SE8 oligonucleotide onlyweakly disrupts binding (compare lane 3 with lane 1).

The previously published DEAF-1-binding site, TTCG, isbased on footprint assays using the Dfd autoregulatory enhancerand the 1.28 gene enhancer (15, 16). This analysis was extendedby performing systematic evolution of ligands by exponential(SELEX) enrichment experiments (30). Recombinant His-tagged DEAF-1 protein was incubated with a random library of

Fig. 1. Mutation of the Mtk regulatory sequence reveals functional motifs. (A) Diagram of the 208-bp Mtk regulatory sequence. Known Rel and GATAfactor-binding sites are boxed in gray. The transcription start site determined by 5� RACE (FirstChoice RLM-RACE kit; Ambion) is to the right of the arrow. Elementsselected for mutation (E1–E11) are underlined. (B) The region in A bearing no alterations (WT) or bearing scrambled elements (SE1–SE11) was cloned into aluciferase reporter construct that was then transiently transfected into S2 cells. After a 6-h induction with LPS (10 �g/ml O127:B8; Sigma), cells were harvestedand luciferase activity was measured. Error bars correspond to one standard deviation of the mean (SDM). (C ) Transgenic larvae carrying Mtk-lacZ (Upper) orMtk SE8-lacZ (Lower) constructs were generated via P-element insertion. Third instar transgenic larvae were infected with a mixture of E. coli and M. luteus andstained for lacZ 6 h later. Shown is a closeup of the fat body. (D) Transgenic Mtk-lacZ (Upper) and Mtk SE8-lacZ (Lower) larvae were fed E. carotovora and dissected6 h later. The intact transgene produces constitutive activity in the proventriculus and inducible activity in the anterior midgut. The transgene carrying the SE8mutation shows no detectable activity in either tissue (Lower).

8352 � www.pnas.org�cgi�doi�10.1073�pnas.0802921105 Reed et al.

radiolabeled oligonucleotides. Protein–DNA complexes weregel-purified and PCR-amplified, and the selected DNA wassubjected to two additional rounds of binding and amplification.The DNAs were sequenced and aligned with the publishedfootprint data (16) to generate a position-weighted matrix [Fig.2D and supporting information (SI) Fig. S1]. The broadestconsensus sequence using this approach is TTCGGBT. TheSELEX data show a weaker preference for cytosine at position3 than the previous footprinting data and a stronger selection forguanine at position 5. This finding may reflect differences in thebinding of DEAF-1 to individual sites, compared with theclustered sites seen in the Dfd enhancer.

The strongest selection by DEAF-1 occurs at positions 2–5(TCGG). We verified the importance of each position by perform-ing gel shift competition assays (Fig. 2E). DEAF-1–E8 complexeswere incubated with unlabeled oligonucleotides bearing mutationsat each position along the consensus binding sequence. Oligonu-cleotides bearing mutations at position 1, 6, or 7 (TTCGGBT)successfully competed with radiolabeled E8 for DEAF-1 binding,suggesting that they contain an intact core-binding sequence (com-pare lanes 5–7 and 21–26 with lane 1). In contrast, mutations atpositions 2–5 (TTCGGBT) greatly impaired competitive binding ofthe modified oligonucleotides (compare lanes 8–20 with lane 1).Hence, strong DEAF-1-binding sites appear to contain a TCGGcore sequence.

DEAF-1 Activates Mtk Expression. Transient transfection assayswere done to investigate the ability of DEAF-1 to activatetranscription in S2 cells. DEAF-1 was expressed in S2 cells byplacing the DEAF-1 coding sequence under the control of theactin promoter (pMA6-DEAF-1). This DEAF-1 expression vec-tor was cotransfected with an Mtk-Luc reporter construct in S2cells (Fig. 3A). The fusion gene is normally induced 12-fold uponaddition of LPS to the culture medium. The addition of pMA6-DEAF-1 causes a 5-fold increase in the basal expression of the

Mtk-Luc reporter gene and a 28-fold increase upon addition ofLPS. In contrast, an Mtk-Luc reporter construct bearing thescrambled E8 sequence (Mtk SE8-Luc) did not respond toexpression of DEAF-1.

We surveyed other immunity genes for sequences that con-formed to the DEAF-1-binding consensus. The 746-bp 5� enhancerof Drs (6) contains five potential DEAF-1-binding sites (E8.1–E8.5)(Fig. 3B). Each site binds DEAF-1 with a different affinity (Fig.3C); sites E8.3 and E8.4 bind particularly well. The five recognitionsequences were scrambled in the context of an otherwise normalDrs enhancer and analyzed in S2 cells by using a Drs-Luc reporterconstruct (Fig. 3B). A Drs-Luc reporter containing wild-typesequences is induced �2-fold upon LPS addition. Cotransfectionwith the pMA6-DEAF-1 expression vector causes an additional12-fold increase in Drs-Luc reporter gene expression in the absenceof LPS and a 17-fold increase with LPS. Mutations in individualbinding sites reduce both basal and activated transcription. Muta-tion of site 8.3 causes the most dramatic reduction in activity.

The 201-bp Diptericin enhancer mediates strong expression inthe fat bodies of infected larvae, but is only weakly induced in S2cells (�2-fold) (Fig. 3D). The enhancer likely lacks DEAF-1-binding sites based on DNA sequence analysis, and a Dpt-Lucreporter construct responds only weakly to transfected DEAF-1 inS2 cells. To determine whether insertion of a DEAF-1 consensusbinding site can confer an ability to respond to DEAF-1 expression,we placed a single, optimal DEAF-1 site 10 bp upstream of theendogenous GATA motif, either in the forward or reverse orien-tation. Both constructs exhibit a significant increase in luciferaseexpression when coexpressed with a DEAF-1 expression construct,compared with the unmodified reporter fusion gene. In contrast,insertion of the scrambled E8 (SE8) sequence led to only a modestincrease in reporter gene expression.

Deaf-1 Synergizes with Dorsal, Dif, and Relish. Mutation of all threeRel sites or all three GATA sites within the Mtk enhancer causes

Fig. 2. Element 8 binds DEAF-1. (A) Schematic of the pulldown assay. Wild-type and SE8 regulatory sequences were PCR-amplified by using a 5� biotinylatedprimer and coupled to magnetic Dynal beads. An EcoRI site was included between the biotin moiety and Mtk sequences. The bead–DNA complexes were blockedwith BSA and incubated with nuclear extracts from S2 cells induced with LPS. After extensive washing, the nucleoprotein complexes were eluted with an EcoRIdigest and subjected to MudPIT analysis. (B) Recombinant His-tagged DEAF-1 protein was incubated with radiolabeled E8 (CATTCATTCGGCTGC) and SE8(CATACGTGTCCTTGC) oligos, and complexes were resolved on 6% native acrylamide gels. Protein amounts range from 10–300 ng. (C) Briefly, 200 ng of DEAF-1was incubated with 6 pmol of radiolabeled E8 oligo (lanes 1–5) and then challenged with a 3� or 10� excess of cold SE8 (lanes 2 and 3) or E8 (lanes 4 and 5)oligos. (D) Position weighted matrix for DEAF-1 derived from published footprinting data (16) and SELEX assays (see Fig. S1). (E) DEAF-1 protein was bound toradiolabeled E8 oligo and challenged with a 3�, 10�, or 30� excess of cold oligos bearing single changes in the DEAF-1 consensus indicated in D.

Reed et al. PNAS � June 17, 2008 � vol. 105 � no. 24 � 8353

IMM

UN

OLO

GY

a complete loss of induced activity in S2 cells (6). Thus, Rel andGATA factors function synergistically to activate immunity geneexpression. The presence of DEAF-1-binding sites near the Reland GATA sites suggests that it may cooperate with these factorsduring the mounting of an immune response. To test this wecotransfected pMA6-DEAF-1 with expression vectors for Dor-sal, Dif, and Relish. Both Mtk-Luc and Drs-Luc fusion geneswere used as reporters to monitor the combinatorial activities ofthese transcription factors (Fig. 4 A–F).

The Mtk-Luc reporter gene (Fig. 4 A–C) is induced 12-foldwith LPS. Addition of 1 �g of pMA6-DEAF-1 raised the basalactivity 5-fold and boosts the response to added LPS �30-fold.Separate transfections with individual Dorsal (Fig. 4A), Dif (Fig.4B), and Relish (Fig. 4C) expression vectors also result insubstantial activation (up to 67-fold for Dorsal, 166-fold for Dif,and 350-fold for Relish with LPS). Cotransfection of pMA6-DEAF-1 with each Rel factor results in some degree of synergy,denoted in italics above each bar graph. The level of synergy iscalculated by dividing the activity of two factors working to-gether by the sum of their individual activities. The average foldsynergy with DEAF-1 on the Mtk-Luc reporter is 1.6-fold withDorsal, 2.2-fold with Dif, and 1.7-fold with Relish.

Similar results were obtained by using the Drs-Luc reporter (Fig.4 D–F), which is induced an average of 2.8-fold upon addition ofLPS in S2 cells. The addition of 1 �g of pMA6-DEAF-1 raises basalactivity 6-fold and LPS induction 10-fold. Separate transfections ofeither Dorsal or Dif dramatically activate this reporter (up to60-fold with LPS), whereas Relish appears to be a much weakeractivator (7-fold with LPS). Cotransfection of pMA6-DEAF-1 witheach Rel factor results in an average of 3-fold synergy with Dif and1.9-fold synergy with Dorsal. Only additive effects (average 1.2-foldsynergy) were obtained with Relish. Altogether, these resultssuggest that DEAF-1 differentially augments the activities of dif-

ferent Rel factors during the induction of immunity genes. Partic-ularly strong synergy is seen between DEAF-1 and Dif.

In summary, we have presented evidence that DEAF-1 is anessential component of the immune response in both Drosophilalarvae and S2 cells. It appears to augment the synergisticactivities of Rel and GATA transcription factors during theimmune response. In so doing, it provides a signal amplificationmechanism so that certain innate immunity genes, such as Mtk,can be transcribed at particularly high levels while using the samesignaling pathways as other less highly expressed immunitygenes. The critical E8 motif, TTCGGCT, is highly conservedamong the Mtk enhancers of divergent Drosophilids and isclosely linked to a paired set of Rel and GATA sites. It istherefore conceivable that DEAF-1 facilitates the binding ortranscriptional efficacy of Rel and GATA factors at linked sites.

The requirement for DEAF-1 in the regulation of Drs, but notDpt, hints at a possible function for DEAF-1 in Toll signaling.Cotransfection experiments using Mtk-Luc and Drs-Luc reportergenes demonstrate that DEAF-1 synergizes with Dorsal andespecially Dif, two effectors of Toll signaling. Only weak coop-eration occurs between DEAF-1 and Relish, a target of the Imdpathway. Microarray studies of flies mutant for Toll and Imdpathway components (31) have comprehensively identifiedgroups of genes coregulated by each pathway. Interestingly,several genes that require Toll signaling for regulation, such asCactus, IM1, and Dif, contain one perfect and several near-perfect consensus DEAF-1-binding sites within 1 kb of thetranscription start sites. Future studies should determinewhether DEAF-1 is a constitutive component of immune tissueslike the GATA factors or is regulated in response to infection byToll signaling as seen for the Rel factors.

MethodsSite-Directed Mutagenesis. Primers carrying the desired mutation plus 15–20 ntflanking each end were phosphorylated in 10 mM Tris (pH 8.0), 1� T4 ligase

Fig. 3. DEAF-1 stimulates transcription of immunity genes. (A) Diagram of the 208-bp Mtk regulatory sequence with approximate location of Rel- (red), GATA-(green), and DEAF-1 (blue)-binding sites is noted. The red diamond denotes a bidirectional Rel site. The striped triangle represents an additional consensus DEAF-1site that overlaps the second GATA site. The effect of cotransfecting pMA6-DEAF-1 (2 �g) with either wild-type or SE8 templates is shown in the bar graph. (B)Diagram of a 746-bp Drs regulatory sequence containing close matches to the DEAF-1 consensus. The Drs transcription start site was mapped by using 5� RACE(FirstChoice RLM-RACE kit; Ambion) to the sequence CCAAGCCACAAGTCG (starting nucleotide underlined). One DEAF-1 site (striped triangle) overlaps a Rel site.The five nonoverlapping DEAF-1 sites (E8.1–E8.5) were mutagenized and tested for a response to transfected pMA6-DEAF-1 (2 �g). (C ) Gel shift showing DEAF-1binding to normal and scrambled recognition elements from B. (D) The E8 motif from Mtk was inserted into the Dpt regulatory region in either a forward (FW)or reverse (RV) orientation relative to the promoter. For comparison, the SE8 sequence also was inserted (white triangle). Error bars correspond to one SDM.

8354 � www.pnas.org�cgi�doi�10.1073�pnas.0802921105 Reed et al.

buffer (New England BioLabs), 1 unit/�l polynucleotide kinase (New EnglandBioLabs), and 5 pmol/�l primer for 1 h at 37°C. Then 10 pmol of phosphory-lated primer was used in a PCR containing 50 ng template, 1� Pfu Turbo buffer(Stratagene), 0.2 mM dNTP mix, and 0.1 unit/�l Pfu Turbo polymerase (Strat-agene). Thermocycling parameters were 1 min at 94°C, 30 cycles at 94°C for 1min, 55°C for 1 min, 65°C for 10 min, ending at 15°C. PCR products wereDpnI-digested (1.6 units/�l; NEB) for 1.5 h, transformed into DH5� cells, andscreened for mutations by sequencing. The scrambled Mtk sequences indi-cated in Fig. 1A are as follows: Mtk SE1, GCAGCG; Mtk SE2, GGTTTG; Mtk SE3,GCGTTGG; Mtk SE4, ACGCTAGCA; Mtk SE5, CGCTCGCC; Mtk SE6, TCAGCTGA;Mtk SE7, TCGCATCC; Mtk SE8, ACGTGTCCT; Mtk SE9, AGACAATCAG; MtkSE10, CGTCGACTA; and Mtk SE11, TGTCTGCTGT. The wild-type and scrambledDrs sequences are as follows: E8.1/SE8.1, ATCGGTG/CTAGGTG; E8.2/SE8.2,TTCGGTA/GACTGTT; E8.3/SE8.3, TACCGAA/CAGTAAC; E8.4/SE8.4, ACCCGAC/CACGACC; and E8.5/SE8.5, ATCGGCT/ACGTTCG.

Affinity Chromatography and MudPIT. The wild-type or SE8 Mtk regulatoryregion from �208 to �33 was PCR-amplified with a biotinylated 5� primer(ELIM Biopharmaceuticals) harboring an EcoRI site (5�-biotin-AAAAGAAT-TCTAGGCTGATAATCCGGGACCGTGGGAAGTC-3�) and unlabeled 3� primer(5�-TGCATCTTAGCTCGGTGGCGGGAATTGATTG). Then 40 pmol PCR productwas coupled to 2 mg of Dynabeads (Dynal Biotech) per the manufacturer’sinstructions. DNA–bead complexes were blocked for 1 h at 4°C in BC-100buffer [20 mM Hepes (pH 7.9), 20% glycerol, 100 mM KCl, 0.2 mM EDTA (pH8.0), 0.5 mM PMSF, 0.5 mM DTT] plus 5% BSA, 0.1 mg/ml salmon sperm DNA,and 0.01% Triton X-100. Nuclear extracts (1 mg) from S2 cells induced 6 h withLPS (10 �g/ml O127: B8; Sigma) were added and incubated for 2 h at 4°C.Protein–DNA complexes were washed one time with blocking buffer and sixtimes with BC-100 containing no PMSF, plus 0.1 mg/ml salmon sperm DNA and0.01% Triton X-100. Complexes were eluted from the Dynal beads with a10-min EcoRI digest (1 unit/�l FPLC pure; Amersham) at 37°C. Eluted protein

complexes were precipitated by the addition of trichloroacetic acid and thendigested by the sequential addition of lys-C and trypsin proteases as describedpreviously (32). The digested samples were then separated by using an onlinemultidimensional seven-step chromatographic strategy, followed by tandemmass spectrometric analysis of the fractionated peptides as they eluted di-rectly into a LTQ-Orbitrap mass spectrometer (Thermofisher) (33). Data anal-ysis was performed by using the SEQUEST and DTASelect2 algorithms, andpeptide identifications were filtered by using a false-positive rate of �5% asestimated by using a decoy database strategy (34–36).

Fly Strains and Infections. All flies were maintained at 25°C on standardcornmeal medium. The method for creating transgenic lines and causingseptic infection was described in ref. 6. The method for E. carotovorainfection was described in ref. 9. At least three independent lines wereanalyzed per experiment.

Cell Culture and Transfections. S2 cells (Invitrogen) were maintained at 25°C inSchneider’s Insect Media supplemented with 10% heat-inactivated FBS and1� Pen/Strep (Sigma). For transfections, cells were seeded in 24-well plates ata density of 2 � 106 cells per ml. One day after plating, transfections werecarried out by using the calcium phosphate method (37). Two days aftertransfection, 10 �g of LPS (O127:B8; Sigma) was added, and the cells wereharvested 6 h later. Cells were lysed (1� reporter lysis buffer; Promega), andluciferase activity was measured on a TD-20/20 luminometer (Turner Designs).The amount of reporter DNA was 200 ng in each experiment.

SELEX and EMSA assays. SELEX and EMSA assays were performed as describedpreviously (6).

ACKNOWLEDGMENTS. This work was supported by National Institutes ofHealth Grant GM46638 (to M.S.L.).

1. Engstrom Y, Kadalayil L, Sun SC, Samakovlis C, Hultmark D, Faye I (1993) kappaB-like motifs regulate the induction of immune genes in Drosophila. J Mol Biol232:327–333.

2. Han ZS, Ip YT (1999) Interaction and specificity of Rel-related proteins in regulatingDrosophila immunity gene expression. J Biol Chem 274:21355–21361.

3. Kadalayil L, Petersen UM, Engstrom Y (1997) Adjacent GATA and kappa B-like motifsregulate the expression of a Drosophila immune gene. Nuc Acids Res 25:1233–1239.

4. Kappler C, Meister M, Lagueux M, Gateff E, Hoffmann JA, Reichhart JM (1993) Insectimmunity. Two 17 bp repeats nesting a kappa B-related sequence confer inducibility tothe diptericin gene and bind a polypeptide in bacteria-challenged Drosophila. EMBOJ 12:1561–1568.

5. Petersen UM, et al. (1999) Serpent regulates Drosophila immunity genesin the larval fat body through an essential GATA motif. EMBO J 18:4013–4022.

Fig. 4. Synergy between DEAF-1 and Rel factors. (A–C) The Mtk-Luc fusion gene was transfected with pMA6-DEAF-1 in the absence or presence of increasingamounts of pMA6-Dorsal (A), pMA6-Dif (B), or pMA6-Relish (C) both with and without the addition of LPS as noted. The average of three experiments is shown.Fold synergy is shown in italics. (D–F) Same as A–C, except the Drs-Luc fusion gene is monitored for reporter expression. Error bars correspond to one SDM.

Reed et al. PNAS � June 17, 2008 � vol. 105 � no. 24 � 8355

IMM

UN

OLO

GY

6. Senger K, et al. (2004) Immunity regulatory DNAs share common organizationalfeatures in Drosophila. Mol Cell 13:19–32.

7. Rehorn KP, Thelen H, Michelson AM, Reuter R (1996) A molecular aspect of hemato-poiesis and endoderm development common to vertebrates and Drosophila. Devel-opment 122:4023–4031.

8. Sam S, Leise W, Hoshizaki DK (1996) The serpent gene is necessary for progressionthrough the early stages of fat-body development. Mech Dev 60:197–205.

9. Senger K, Harris K, Levine M (2006) GATA factors participate in tissue-specific immuneresponses in Drosophila larvae. Proc Natl Acad Sci USA 103:15957–15962.

10. Dushay MS, Asling B, Hultmark D (1996) Origins of immunity: Relish, a compoundRel-like gene in the antibacterial defense of Drosophila. Proc Natl Acad Sci USA93:10343–10347.

11. Ip YT, et al. (1993) Dif, a dorsal-related gene that mediates an immune response inDrosophila. Cell 75:753–763.

12. Reichhart JM, et al. (1993) Expression and nuclear translocation of the rel/NF-kappaB-related morphogen dorsal during the immune response of Drosophila. C R Acad SciIII 316:1218–1224.

13. Uvell H, Engstrom Y (2007) A multilayered defense against infection: Combinatorialcontrol of insect immune genes. Trends Genet 23:342–349.

14. Uvell H, Engstrom Y (2003) Functional characterization of a novel promoter elementrequired for an innate immune response in Drosophila. Mol Cell Biol 23:8272–8281.

15. Gross CT, McGinnis W (1996) DEAF-1, a novel protein that binds an essential region ina Deformed response element. EMBO J 15:1961–1970.

16. Pederson JA, et al. (2000) Regulation by homeoproteins: A comparison of Deformed-responsive elements. Genetics 156:677–686.

17. Li X, Veraksa A, McGinnis W (1999) A sequence motif distinct from Hox binding sitescontrols the specificity of a Hox response element. Development 126:5581–5589.

18. Gibson TJ, Ramu C, Gemund C, Aasland R (1998) The APECED polyglandular autoim-mune syndrome protein, AIRE-1, contains the SAND domain and is probably a tran-scription factor. Trends Biochem Sci 23:242–244.

19. Surdo PL, Bottomley MJ, Sattler M, Scheffzek K (2003) Crystal structure and nuclearmagnetic resonance analyses of the SAND domain from glucocorticoid modulatoryelement binding protein-1 reveals deoxyribonucleic acid and zinc binding regions. MolEndocrinol 17:1283–1295.

20. Bottomley MJ, et al. (2001) The SAND domain structure defines a novel DNA-bindingfold in transcriptional regulation. Nat Struc Biol 8:626–633.

21. Veraksa A, Kennison J, McGinnis W (2002) DEAF-1 function is essential for the earlyembryonic development of Drosophila. Genesis 33:67–76.

22. Huggenvik JI, et al. (1998) Characterization of a nuclear deformed epidermal auto-regulatory factor-1 (DEAF-1)-related (NUDR) transcriptional regulator protein. MolEndocrinol 12:1619–1639.

23. LeBoeuf RD, et al. (1998) Molecular cloning, sequence analysis, and tissue distributionof suppressin, a novel suppressor of cell cycle entry. J Biol Chem 273:361–368.

24. Michelson RJ, et al. (1999) Nuclear DEAF-1 related (NUDR) protein contains a novel DNAbinding domain and represses transcription of the heterogeneous nuclear ribonucle-oprotein A2/B1 promoter. J Biol Chem 274:30510–30519.

25. LeBoeuf RD, Burns JN, Bost KL, Blalock JE (1990) Isolation, purification, and partialcharacterization of suppressin, a novel inhibitor of cell proliferation. J Biol Chem265:158–165.

26. Carr DJ, Blalock JE, Green MM, LeBoeuf RD (1990) Immunomodulatory characteristicsof a novel antiproliferative protein, suppressin. J Neuroimmunol 30:179–187.

27. Leulier F, et al. (2003) The Drosophila immune system detects bacteria through specificpeptidoglycan recognition. Nat Immunol 4:478–484.

28. Kaneko T, et al. (2004) Monomeric and polymeric gram-negative peptidoglycanbut not purified LPS stimulate the Drosophila IMD pathway. Immunity 20:637–649.

29. Wolters DA, Washburn MP, Yates JR, III (2001) An automated multidimensional proteinidentification technology for shotgun proteomics. Anal Chem 73:5683–5690.

30. Gold L, Polisky B, Uhlenbeck O, Yarus M (1995) Diversity of oligonucleotide functions.Annu Rev Biochem 64:763–797.

31. De Gregorio E, Spellman PT, Tzou P, Rubin GM, Lemaitre B (2002) The Toll and Imdpathways are the major regulators of the immune response in Drosophila. EMBO J21:2568–2579.

32. McDonald WH, Ohi R, Miyamoto DT, Mitchison TJ, Yates JR III (2002) Comparison ofthree directly coupled HPLC MS/MS strategies for identification of proteins fromcomplex mixtures: Single-dimension LC-MS/MS, 2-phase MudPIT, and 3-phase MudPIT.Int J Mass Spectrom 219:245–251.

33. MacCoss MJ, et al. (2002) Shotgun identification of protein modifications from proteincomplexes and lens tissue. Proc Natl Acad Sci USA 99:7900–7905.

34. Eng JK, McCormack AL, Yates JR, III (1994) An approach to correlate tandem massspectral data of peptides with amino acid sequences in a protein database. J Am SocMass Spectrom 5:976–989.

35. Peng J, Elias JE, Thoreen CC, Licklider LJ, Gygi SP (2002) Evaluation of multidimensionalchromatography coupled with tandem mass spectrometry (LC/LC-MS/MS) for large-scale protein analysis: The yeast proteome. J Proteome Res 2:43–50.

36. Tabb DL, McDonald WH, Yates JR III (2002) DTASelect and Contrast: Tools forassembling and comparing protein identifications from shotgun proteomics. JProteome Res 1:21–26.

37. Rio DC, Rubin GM (1985) Transformation of cultured Drosophila melanogaster cellswith a dominant selectable marker. Mol Cell Biol 5:1833–1838.

8356 � www.pnas.org�cgi�doi�10.1073�pnas.0802921105 Reed et al.