Multiple enhancers ensure precision of gap -expression patterns in the Drosophila

Michael W. Perrya, Alistair N. Boettigerb, and Michael Levinec,d,1

aDepartment of Integrative Biology, bBiophysics Graduate Group, cCenter for Integrative Genomics, Division of Genetics, Genomics, and Development, and dDepartment of Molecular and Cell Biology, University of California, Berkeley, CA 94720-3200

Contributed by Michael Levine, June 22, 2011 (sent for review May 25, 2011) of the Drosophila embryo begins with the establish- discuss potential mechanisms for such synergy and ment of spatially restricted gap gene-expression patterns in re- suggest that minimal enhancers producing aberrant patterns of sponse to broad gradients of maternal transcription factors, such might nonetheless contribute to authentic ex- as Bicoid. Numerous studies have documented the fidelity of these pression profiles in the context of their native loci. expression patterns, even when are subjected to genetic or environmental stress, but the underlying mechanisms for this Results and Discussion transcriptional precision are uncertain. Here we present evidence Every Gap Gene Contains Multiple Enhancers for a Single Gap Pattern. that every gap gene contains multiple enhancers with overlapping Candidate gap enhancers were identified using ChIP-chip data activities to produce authentic patterns of gene expression. For (11). Specifically, clustered binding sites for maternal and gap example, a recently identified hunchback (hb) enhancer (located proteins were identified within 100 kb of every gap gene (see 5-kb upstream of the classic enhancer) ensures repression at the Materials and Methods). This survey identified each of the known anterior pole. The combination of intronic and 5′ knirps (kni) en- enhancers, as well as putative shadow enhancers (12–17). For hancers produces a faithful expression pattern, even though the example, a potential distal shadow enhancer was identified for intronic enhancer alone directs an abnormally broad expression hb, located 4.5-kb upstream of the proximal transcription start pattern. We present different models for “enhancer synergy,” site (designated “P2” in ref. 18) and upstream of the later-acting

whereby two enhancers with overlapping activities produce au- distal promoter (designated “P1”) (Fig. 1C). BIOLOGY fi thentic patterns of gene expression. A 400-bp genomic DNA fragment from this newly identi ed DEVELOPMENTAL region was attached to a lacZ reporter gene and expressed in cis-regulation | development | robustness transgenic embryos (Fig. 1B). The resulting hb/lacZ fusion gene exhibits localized expression in anterior regions of the embryo “ ” ecent studies identified “shadow” enhancers for en- similar to that seen for the endogenous gene and classic en- fi Rgaged in the dorsal-ventral patterning of the early Drosophila hancer identi ed over 20 y ago (13, 19) (Fig. 1B; compare with embryo (1). These enhancers are sometimes located within A). The classic proximal and distal shadow enhancers exhibit neighboring genes, and along with conventional, proximal en- similar responses to increasing Bicoid copy number (Fig. S1A). fi hancers, they produce robust patterns of gene expression in early ChIP-chip data also identi ed potential pairs of enhancers for – – embryos under stress (2, 3). For example, the snail gene exhibits Kr (Fig. 1 D F) and kni (Fig. 1 G I). There are two distinct erratic patterns of activation in embryos raised at 30 °C when either clusters of binding sites upstream of Kr. The fi “ ” the proximal or shadow enhancer is removed (3). It was proposed previously identi ed Kr CD2 enhancer contains the proximal “ ” enhancer but also part of the distal binding cluster (14). Sub- that shadow enhancers represent a mechanism of canalization fi (4), whereby populations of embryos develop normally even when sequent lacZ fusion assays identi ed each ChIP-chip peak and underlying binding sites as separable proximal and distal en- subject to variations in temperature or genetic background. – fi In the present study we provide evidence that many of the hancers (Fig. 1 D F). Similarly, more re ned limits were de- termined for the kni intronic enhancer (Fig. 1 G and I, and Fig. genes controlling anterior-posterior patterning contain multiple fi ′ enhancers with overlapping activities, including head patterning S1B) (12), in addition to the previously identi ed 5 distal en- Kr kni genes and gap genes, which initiate the segmentation gene net- hancer (20). Both the distal enhancer and the intronic enhancer produce somewhat broader patterns of expression than work (e.g., refs. 5 and 6). For example, the gap genes hunchback the endogenous gene (Fig. 1 E and G) (see below). Additional (hb), Kruppel (Kr), and knirps (kni) are each regulated by two gap enhancers were also identified for giant, including an addi- distinct enhancers that control the initial bands of gene expression tional distal enhancer located ∼35-kb downstream within a within the presumptive head, thorax, and abdomen. Evidence is neighboring gene (Fig. S2A and ref. 12). presented that the two enhancers work together (enhancer syn- The survey of gap and maternal binding clusters was extended ergy) to ensure uniform expression within correct spatial limits. to include the so-called “head” and “terminal” gap genes, critical Previous studies have documented examples of enhancer au- for the differentiation of head structures and the nonsegmented tonomy and enhancer interference. Multiple enhancers often termini of early embryos (Fig. 1 J–O and Fig. S2 B–J). Additional produce additive patterns of gene expression, as seen for the 7- fi fi enhancers were identi ed for empty-spiracles (ems) (original stripe even skipped (eve) expression pattern arising from ve sep- enhancer identified in ref. 21) (Fig. 1 M–O), huckebein (hkb) arate enhancers (two located 5′ of the eve transcription unit and – – (original enhancer in ref. 22) (Fig. S2 E G), and forkhead (fkh) three located downstream of the gene) (7 9). Sometimes, multiple (original enhancer in ref. 12) (Fig. S2 B–D). More refined limits enhancers interfere with one another when placed within a com- mon regulatory region. For example, ventral repressors that de- lineate the intermediate neuroblasts defective (ind) expression Author contributions: M.W.P., A.N.B., and M.L. designed research; M.W.P. and A.N.B. pattern block the activities of a neighboring eve stripe-3 enhancer, performed research; M.W.P. and A.N.B. contributed new reagents/analytic tools; and conversely, repressors that establish the posterior limit of M.W.P., A.N.B., and M.L. analyzed data; and M.W.P., A.N.B., and M.L. wrote the paper. the stripe-3 pattern interfere with ind (10). The authors declare no conflict of interest. In the present study, evidence is presented that combining 1To whom correspondence should be addressed. E-mail: [email protected]. multiple enhancers in a common regulatory region can produce This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. sharper and more homogeneous patterns of gene expression. We 1073/pnas.1109873108/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1109873108 PNAS Early Edition | 1of6 Downloaded by guest on September 23, 2021 1000 bp hb proximal ABhb distal C

hb P1 CG8112

hb P2 Kr proximal Kr distal 1000 bp DEKr F

kni proximal GHkni distal 1000 bp I

kni

oc/otd JKoc/otd distal L 1000 bp

otd

ems ems distal MN O 1000 bp ems

= proximal = distal

Fig. 1. Activities of gap enhancers identified by in situ hybridization. (A and B) The hb/lacZ transgenes containing the (A) proximal (classic) or (B) newly identified distal enhancer. The locations of these enhancers are shown in C.(D and E) Kr/lacZ transgenes containing the (D) proximal or (E) distal enhancer. The locations of these enhancers are shown in F.(G and H) The kni/lacZ transgenes containing either the (G) proximal intronic enhancer or (H) the distal 5′ enhancer. The locations of these enhancers are shown in I.(J and K) Expression of endogenous oc/otd (J); oc/lacZ transgene containing an intronic enhancer (K). The locations of the oc/otd enhancers are shown in L.(M and N) Expression of endogenous ems (M); ems/lacZ transgene containing a distal enhancer (N). Locations of the enhancers shown in O.

were also determined for the previously identified ocelliless/ faithful compared with the BAC transgene containing both orthodenticle (oc/otd) intronic enhancer (12) (Fig. 1 J–L). For enhancers (Fig. 2C). Embryos were double-labeled to detect simplicity, we will hereafter refer to the two enhancers regulating both yellow and hb nascent transcripts. During nuclear cleavage a given gap gene as proximal and distal, based on their relative cycle (cc) 13, a substantial fraction of nuclei (14%) expressing locations to the transcription start site. hb nascent transcripts lack yellow transcription upon removal of the shadow enhancer (Fig. 2A). An even higher fraction of hb Multiple Enhancers Produce Authentic Expression Patterns. BAC nuclei (24%) lack yellow transcription when the proximal en- recombineering (23, 24), phiC31-targeted genome integration hancer is removed (Fig. 2B). Control transgenic embryos con- (25, 26), and quantitative in situ hybridization assays (3, 27) were taining both enhancers exhibit more uniform patterns of trans- used to determine the contributions of the proximal and distal cription, whereby only an average of ∼3% of nuclei fail to match enhancers to the hb expression pattern (Fig. 2). BACs containing ∼ fl the endogenous pattern of transcription (Fig. 2C). The distri- 20 kb of genomic DNA encompassing the hb gene and anking “ ” sequences were integrated into the same position in the Dro- bution of missing nuclei across the population of cc13 embryos sophila genome. The hb transcription unit was replaced with the is plotted in Fig. 2D. The pairwise Wilcoxon rank sum test (also called the Mann- yellow gene, which permits quantitative detection of nascent fi transcripts using an intronic hybridization probe (3) (Materials Whitney u test) was used to determine the signi cance of the and Methods and Fig. S3). The modified BAC retains the com- apparent variation in gene expression resulting from the re- plete hb 5′ and 3′ UTRs. Additional BACs were created by in- moval of either the proximal or distal enhancer (Fig. 2D). activating the proximal or distal enhancers by substituting critical Control embryos containing the hb BAC transgene with both regulatory elements with “random” DNA sequences (see diagrams enhancers exhibit some variation in the number of nuclei that above panels in Fig. 2 A–C and Materials and Methods). lack yellow nascent transcripts. Despite this variation, the sta- BAC transgenes lacking either the distal (Fig. 2A) or proximal tistical analyses indicate that the loss of either the proximal or (Fig. 2B) enhancer continue to produce localized patterns of distal enhancer results in a significant change in yellow tran- transcription in anterior regions of transgenic embryos in re- scription patterns compared with the control BAC transgene sponse to the Bicoid gradient. However, the patterns are not as (P =4E-8).

2of6 | www.pnas.org/cgi/doi/10.1073/pnas.1109873108 Perry et al. Downloaded by guest on September 23, 2021 AB

CD BIOLOGY DEVELOPMENTAL

Fig. 2. Function of hb enhancers via BAC transgenesis. An ∼20-kb BAC containing genomic DNA encompassing the hb locus was modified to remove specific proximal or distal enhancers. The hb transcription unit was replaced with the yellow reporter gene to identify de novo transcripts via in situ hybridization using a probe directed against the yellow intron (see diagram above images in A–C). (A) hb-BAC with distal enhancer inactivated (see X in diagram), (B) hb- BAC with proximal enhancer inactivated, and (C) hb-BAC with both enhancers intact. The median ratio of “discordance” is indicated beneath each image. This is the fraction of nuclei that express endogenous hb nascent transcripts, but not yellow transcripts. (D) Cumulative frequency distributions for the fraction of “missing nuclei” in the three populations of embryos. The ordinate axis gives the probability of observing an embryo from this population, with fewer than the abscissa fraction of nuclei transcribing the endogenous gene but not the reporter. Statistical comparisons between the distributions are presented above the panel, with subscripts matching the panel labels [i.e., pCA is the P value from the pair wise comparison of the distribution of embryos with the hb control BAC, (C), to those with the distal enhancer removed (A)].

Distal hb Enhancer Mediates Dominant Repression. The preceding (Fig. 3 B and C). It should be noted that the BAC transgene analyses suggest that multiple enhancers produce more uniform lacking the proximal enhancer exhibits “super-repression” be- patterns of de novo transcription than individual proximal or cause of reduced activation at the anterior pole (P = 0.012) distal enhancers. Additional studies were done to determine (Fig. 3D). whether multiple enhancers also help produce authentic spatial These observations suggest that the distal enhancer contains limits of transcription (Fig. 3). repression elements that function in a dominant manner to at- The expression of hb normally diminishes at the anterior pole tenuate the activities of the proximal enhancer at the anterior of cc13 to 14 embryos. This loss in expression has been attributed pole. There is a loss of repression when the distal enhancer to attenuation of Bcd activity by Torso RTK signaling (e.g., ref. driving lacZ is crossed into a torso mutant background (Fig. S4). 28). However, the proximal enhancer fails to recapitulate this This observation implicates one or more repressors functioning loss (29) (Fig. 1A). In contrast, the distal enhancer is inactive at downstream of Torso signaling, including Tailless and Huck- the anterior pole (Fig. 1B), and the two enhancers together ebein, which have been shown to function as long-range domi- produce a pattern that is similar to endogenous expression, in- nant repressors (30–32). Indeed, whole-genome ChIP assays cluding reduced expression at the pole (Fig. 2C). identify more potential Tll and Hkb binding sites in the distal vs. To examine the relative contributions of the proximal and proximal enhancer (11) (Fig. S5A). The persistence of hb ex- distal enhancers in this repression, yellow nascent transcripts pression in anterior regions has been shown to be detrimental, were measured in transgenic embryos expressing BAC reporter causing defects in mouth parts and malformation of the gut (33). genes containing one or both hb enhancers (Fig. 2). Particular efforts focused on the early phases of cc14, when repression of Correction of the kni Expression Pattern. Kr/lacZ and kni/lacZ fusion endogenous hb transcripts is clearly evident (Fig. 3). For the genes containing either one or two enhancers were inserted into transgene lacking the proximal, classic enhancer, but containing the same position in the Drosophila genome (Fig. 3). Transgenic the newly identified distal enhancer, a median of 6% (std 6%) of embryos were double-labeled to detect the expression of the nuclei exhibit expression of yellow nascent transcripts but lack transgene (lacZ) as well as the endogenous gap gene. expression of the endogenous gene (Fig. 3A). In contrast, a me- The kni proximal (intronic) enhancer alone produces an ab- dian of 24% (std 11%) of nuclei displays a similar discordance normally broad pattern of expression, especially in posterior upon removal of the distal enhancer. In control embryos, 16% regions (Fig. 3F; but see also Fig. 1G and ref. 12). In contrast, the (std 11%) of nuclei express yellow but lack hb nascent transcripts kni distal (5′) enhancer produces erratic lacZ activation within

Perry et al. PNAS Early Edition | 3of6 Downloaded by guest on September 23, 2021 pairwise Wilcoxon: p = 0.015 p = 0.012 p = 0.00044 hb proximal removed hb distal removed hb control CB CA AB ABCD 1

0.8

0.6

0.4

control 22C N=44 0.2 no distal 22C N=47 cummulative frequency no proximal 22C N=10 σ= σ= σ= 0 6% ( 6%) 24% ( 11%) 16% ( 11%) 0 0.2 0.4 0.6 0.8 1 Transcribe reporter AND endogenous Transcribe reporter but NOT endogenous (anterior pole) fraction of ectopically active nuclei X X

pairwise Wilcoxon: p = 1.8e−06 p = 0.72 p = 2.3e−05 kni distal kni proximal kni 2 enhancer GF GE EF EFGH 1

0.8

0.6

0.4

kni 2 enhancers, 22C N=21 0.2 kni distal, 22C N=28 σ= cummulative frequency kni proximal, 22C N=16 7% (σ=7%) 20% (σ=9%) 7% ( 7%) 0 0 0.2 0.4 0.6 0.8 1 Transcribe reporter AND endogenous Transcribe reporter but NOT endogenous ectopic expression rate

pairwise Wilcoxon: p = 0.88 p = 0.00074 p = 0.00088 KI KJ IJ IJKKr distal Kr proximal Kr 2 enhancer L 1

0.8

0.6

0.4

Kr 2 enhancers, 22C N=17 0.2 Kr distal, 22C N=12 cummulative frequency Kr proximal, 22C N=14 7% (σ=4%) 1% (σ=4%) 0 7% (σ=5%) 0 0.2 0.4 0.6 0.8 1 Transcribe reporter AND endogenous Transcribe reporter but NOT endogenous ectopic expression rate

Fig. 3. Enhancer synergy produces correct spatial limits. Discordance of yellow (A–C)orlacZ (E–G, I–K) transgenes and endogenous gap gene nascent transcripts. Nuclei exhibiting ectopic transgene expression are indicated in red. Sites of concordant expression are indicated in yellow. (A–C) BAC transgenes lacking the proximal (A) or distal (B) enhancer, or containing both enhancers intact (C). Nuclei in the anterior third of the hb-expression region which transcribe the reporter but not endogenous hb are shown in red. (D) Cumulative frequency of nuclei in the anterior third of the hb-expression domain containing yellow, but not endogenous hb, nascent transcripts. Median and SDs are shown on the corresponding panels. (E and F) The kni/lacZ reporter genes driven by (E) distal enhancer, (F) proximal enhancer, or (G) both enhancers. Median fractions of nuclei transcribing lacZ but not the endogenous gene are indicated below each image. (H) Cumulative frequency distributions for the fraction of ectopically active nuclei. (I–L) Similar analysis of Kr/lacZ transgenes containing the distal (I), proximal (J), or both (K) enhancers.

nearly normal spatial limits (Fig. 3E). An essentially normal The modest anterior expansion of the expression pattern pattern of lacZ transcription is observed when both enhancers are driven by the kni intronic enhancer is more difficult to explain combined in a common transgene (intronic enhancer 5′ and distal because this boundary is probably formed by the Hb repressor enhancer 3′ of lacZ) (Fig. 3G). It appears that lacZ transcription (37), which is not known to function in a long-range and domi- is slightly broader than the endogenous pattern, but considerably nant manner. If the action of short-range repressors is also af- narrower than the pattern observed for the intronic enhancer fected by stochastic processes (e.g., binding of the repressor to alone (Fig. 3J)(P = 1.8E-6), and not statistically different from enhancer or looping of a bound enhancer to promoter), perhaps the expression limits of the distal enhancer alone (P = 0.72) (Fig. having two enhancers might improve the chances of maintaining 3L). There is no significant narrowing of the Kr/lacZ expression proper repression. pattern when both the distal and proximal enhancers are com- We have presented evidence that the robust and tightly de- bined within the same transgene (Fig. 3 K and L)(P = 1.0). fined patterns of gap gene expression do not arise from the Perhaps additional Kr regulatory elements are required for the unique action of individual enhancers. Rather, these patterns type of narrowing observed for the kni intronic enhancer. Alter- depend on multiple and separable enhancers with similar, but nately, all of these transgenes use the eve basal promoter and it is slightly distinct regulatory activities. This enhancer synergy pro- possible that promoter-specific interactions are important for duces more homogeneous patterns of transcriptional activity, as establishing the normal limits of the Kr expression pattern. well as more faithful spatial limits of expression. As discussed earlier, long-range repressors bound to the distal The enhancer synergy documented in this study is somewhat hb enhancer might inhibit the activities of the proximal enhancer distinct from the proposed role of the shadow enhancer regu- at the anterior pole of precellular embryos. The distal kni en- lating snail expression in the presumptive mesoderm (3). The hancer might function in a similar manner to sharpen the ex- dual regulation of snail by the proximal and distal (shadow) pression limits of the intronic enhancer. The spatial limits of gap enhancers was shown to ensure homogenous and reproducible gene-expression patterns have been shown to depend on cross- expression in embryo after embryo in large populations of em- repressive interactions (e.g., refs. 34–36). The kni intronic en- bryos, even when they are subject to increases in temperature. In hancer might lack critical gap repression elements because it contrast, dual regulation of hb expression by proximal and distal produces an abnormally broad expression pattern. Indeed, whole- enhancers appears to ensure homogenous activation in response genome ChIP assays identify more putative Tailless binding sites to limiting amounts of the Bicoid gradient. They are used as an in the distal vs. intronic enhancer (11) (Fig. S5B). These Tailless obligatory patterning mechanism rather than buffering environ- repression elements might function in a dominant fashion to mental changes. Despite these apparent differences, it is possible restrict the limits of the intronic enhancer. that dominant repression is also used as a mechanism of synergy

4of6 | www.pnas.org/cgi/doi/10.1073/pnas.1109873108 Perry et al. Downloaded by guest on September 23, 2021 Activators present Repressors present

Enhancer A 10% failure 10% failure C X [R]

1% failure Long range repression

B 10% failure 10%X failure D X X [R]X 10% failure Ectopic expression

Fig. 4. Models for enhancer synergy. (A and B) Activation of one promoter by two enhancers. If two independent enhancers each have a 10% failure rate in activating expression and transcription factor binding or enhancer looping is rate limiting, then the two enhancers should have a combined failure rate of 10% × 10% = 1% (A). Removing one enhancer increases the failure rate to 10% (B). (C and D) The binding of a long range, “dominant” repressor to one enhancer is sufficient to inactivate the other (C). Removal of this enhancer results in “ectopic” expression (D).

for the regulation of snail expression. The distal enhancer con- Draw2 (41). The program and binding motif models used are available on- tains repressor elements (e.g., Huckebein) that inhibit the ex- line at http://line.bioinfolab.net/webgate/submit.cgi. pression of the proximal enhancer at the termini (3). Candidate regions (listed in Table S1) were tested in vivo using traditional Different mechanisms can be envisioned to account for en- lacZ reporter assays combined with targeted phiC31 transgenesis as adapted BIOLOGY for use in Drosophila (25, 26). An nE2G backbone with insulators (42) hancer synergy. Perhaps the simplest is that there are fewer in- DEVELOPMENTAL modified for targeted integration was used to test potential enhancers by active nuclei within a given gap expression domain because of placing them upstream of an eve-lacZ fusion gene. The same construct was the diminished failure rate of successful enhancer-promoter used for the one vs. two enhancer experiments for Kr and kni; the second interactions with two enhancers rather than one. If the rates at enhancer for the two enhancer constructs was added into a BstBI restriction which enhancers fail to activate transcription are completely site downstream of lacZ ∼5-kb away from the first enhancer. The landing independent, then one would expect the combined action of two site 51D (26), Bloomington Stock Center number 24483, was used for enhancers to yield a multiplicative reduction in how often a given lacZ assays. cell fails to express the gene within a given window of time. This The two hb enhancer-lacZ constructs were crossed into a 4× or 6× ma- sort of synergy does not require any direct physical or co- ternal Bicoid copy number background using the BB9+16 fly line (19). operative interactions between the enhancers. Nonetheless, the effect can be significant (as seen for hb). For example, two Recombineering and Transgenesis. Recombineering was performed as de- enhancers, each with a 10% uncorrelated failure rate, may to- scribed previously in ref. 3 (see also refs. 23, 24, 43, and 44). The yellow reporter (used to detect sites of nascent transcript by using an intronic in situ probe) gether be expected to have a 1% failure rate, a 10-fold reduction was integrated as a yellow-kanamycin fusion that left the native hb UTRs in- (Fig. 4 A and B). For genes that produce strong bursts of mRNA tact. The bcd binding-site clusters and surrounding regions of the primary or expression, this change in frequency of transcription may have shadow enhancers were removed via replacement with an ampicillin re- a dramatic effect on the variation of total mRNA levels. sistance cassette taken from pBlueScript. Primers used for construct building A second but critical potential mechanism of enhancer synergy and recombineering are listed in Table S1. BAC CH322-55J23 (24) was the basis concerns long-range, dominant repression. Repressors (such as for all subsequent modifications. All BACs were integrated into landing site Tailless) bound to one enhancer are sufficient to restrict the spatial VK37 on chromosome 2 (23), Bloomington Stock Center number 24872. limits of the other enhancer. There is no need for long-range re- fi pressor elements to appear in both enhancers to achieve normal Whole-Mount in Situ Hybridization. Embryos were xed using standard spatial limits of gene expression. It has been suggested that long- methods. Fluorescent or colormetric in situ hybridization was performed as described in refs. 3 and 45. Probes were generated with the primers listed range repressors, such as Hairy, mediate the assembly of positioned – in Table S1 and in vitro transcription. Reporter genes were labeled with nucleosomes at the core promoter (38 40). Such repressive nu- digoxigenin-tagged antisense probes, sheep anti-dig primary antibodies cleosomes should block productive enhancer–promoter interac- (Roche), and donkey anti-sheep Alexa 555 secondary antibodies (Invitrogen). tions, even for enhancers lacking repressor sites (Fig. 4 C and D). Endogenous genes hb, Kr,andkni were labeled with biotin-tagged probes, Regardless of the detailed molecular mechanisms, the com- mouse anti-bio primary antibodies (Roche), and donkey anti-mouse Alexa bined action of multiple enhancers helps explain why an in- 488 secondaries (Invitrogen). Nuclei were counterstained with DRAQ5 dividual enhancer sometimes fails to recapitulate an authentic (Biostatus Ltd.). expression pattern when taken from its native context. Enhanc- × ers that produce abnormal patterns of expression (e.g., kni Confocal Image Acquisition and Computational Image Processing. The 1,024 1,024 3-color image stacks were acquired using a Leica SL Laser Scanning Con- intronic enhancer) can nonetheless contribute to homogeneous focal microscope as described in ref. 3. Image segmentation and analysis was and robust patterns of gene expression in conjunction with the performed as described in ref. 3, with minor modification. Nuclei were seg- additional enhancers contained within the endogenous locus. mented using a difference-of-Gaussians filter optimized with size selection (Fig. S3D). A space-filling, segment dilation algorithm was used to assign all pixels in Materials and Methods the embryo to one of the segmented nuclei, created a final nuclear mask (Fig. Enhancer Identification and Testing. Prospective enhancers were identified S3E). All nuclear masks were manually checked to confirm accurate segmenta- near genes of interest using a combination of ChIP-chip data [provided for tion. Nascent transcripts were localized for both the reporter and the endoge- various maternal, gap, and pair rule genes by the Berkeley Drosophila nous genes, also using difference-of-Gaussians filters, this time optimized to Transcription Network Project (11)] and sequence-based binding-site cluster detect the bright transcripts corresponding to sites of transcription (Fig. S3). analysis. The cluster analysis was performed using the software Cluster- Intensity thresholds and dot-size thresholds reduced spurious counts. Segmen-

Perry et al. PNAS Early Edition | 5of6 Downloaded by guest on September 23, 2021 tation results were curated by the user. This segmentation enabled the nucleus- on the Github source page for this project on “shadow enhancers”: https:// by-nucleus analysis of transcriptional state of reporter and endogenous gene as github.com/AlistairBoettiger/Shadow-Enhancers. described in the text and shown in Figs. 2 and 3. These analysis scripts were wrapped in a Graphical User Interface implemented through Matlab’s software ACKNOWLEDGMENTS. The authors thank Jacques Bothma, Nipam Patel, GUI Design Environment (GUIDE). The source code for this analysis is available in and Vivek Chopra for helpful suggestions; Chiahao Tsui and Emilia Esposito for technical support; and Steve Small for providing the tll (+4) enhancer line the supplemental material of ref. 3. Updated versions of our image segmen- ’ and Gary Struhl for providing the BB9+16 bcd extra copy line. M.W.P. and tation routines can be found on the authors Github page for image processing A.N.B are the recipients of National Science Foundation predoctoral fellow- tools: https://github.com/AlistairBoettiger/Image_Analysis. All of the source ships. This work was funded by Grant GM34431from the National Institutes codes used to compute and plot the results from this publication are available of Health (to M.L.).

1. Hong J-W, Hendrix DA, Levine MS (2008) Shadow enhancers as a source of evolu- 24. Venken KJ, et al. (2009) Versatile P[acman] BAC libraries for transgenesis studies in tionary novelty. Science 321:1314. . Nat Methods 6:431–434. 2. Frankel N, et al. (2010) Phenotypic robustness conferred by apparently redundant 25. Groth AC, Fish M, Nusse R, Calos MP (2004) Construction of transgenic Drosophila by – transcriptional enhancers. Nature 466:490 493. using the site-specific integrase from phage phiC31. Genetics 166:1775–1782. 3. Perry MW, Boettiger AN, Bothma JP, Levine M (2010) Shadow enhancers foster ro- 26. Bischof J, Maeda RK, Hediger M, Karch F, Basler K (2007) An optimized transgenesis – bustness of Drosophila gastrulation. Curr Biol 20:1562 1567. system for Drosophila using germ-line-specific phiC31 integrases. Proc Natl Acad Sci 4. Waddington CH (1942) Canalization of development and the inheritance of acquired USA 104:3312–3317. characters. Nature 150:563–565. 27. Boettiger AN, Levine M (2009) Synchronous and stochastic patterns of gene activation 5. Clyde DE, et al. (2003) A self-organizing system of repressor gradients establishes in the Drosophila embryo. Science 325:471–473. segmental complexity in Drosophila. Nature 426:849–853. 28. Kim Y, et al. (2010) MAPK substrate competition integrates patterning signals in the 6. Jaeger J, et al. (2004) Dynamic control of positional information in the early Dro- – sophila embryo. Nature 430:368–371. Drosophila embryo. Curr Biol 20:446 451. 7. Small S, Blair A, Levine M (1996) Regulation of two pair-rule stripes by a single en- 29. Struhl G, Johnston P, Lawrence PA (1992) Control of Drosophila body pattern by the – hancer in the Drosophila embryo. Dev Biol 175:314–324. hunchback morphogen gradient. Cell 69:237 249. 8. Small S, Blair A, Levine M (1992) Regulation of even-skipped stripe 2 in the Drosophila 30. Courey AJ, Jia S (2001) Transcriptional repression: The long and the short of it. Genes embryo. EMBO J 11:4047–4057. Dev 15:2786–2796. 9. Fujioka M, Emi-Sarker Y, Yusibova GL, Goto T, Jaynes JB (1999) Analysis of an even- 31. Goldstein RE, Jiménez G, Cook O, Gur D, Paroush Z (1999) Huckebein repressor activity skipped rescue transgene reveals both composite and discrete neuronal and early in Drosophila terminal patterning is mediated by Groucho. Development 126: blastoderm enhancers, and multi-stripe positioning by gap gene repressor gradients. 3747–3755. Development 126:2527–2538. 32. Gray S, Levine M (1996) Short-range transcriptional repressors mediate both 10. Stathopoulos A, Levine M (2005) Localized repressors delineate the neurogenic ec- quenching and direct repression within complex loci in Drosophila. Genes Dev 10: – toderm in the early Drosophila embryo. Dev Biol 280:482 493. 700–710. 11. Li XY, et al. (2008) Transcription factors bind thousands of active and inactive regions 33. Janody F, Reischl J, Dostatni N (2000) Persistence of Hunchback in the terminal region in the Drosophila blastoderm. PLoS Biol 6:e27. of the Drosophila blastoderm embryo impairs anterior development. Development 12. Schroeder MD, et al. (2004) Transcriptional control in the segmentation gene network 127:1573–1582. of Drosophila. PLoS Biol 2:E271. 34. Kraut R, Levine M (1991) Mutually repressive interactions between the gap genes 13. Driever W, Thoma G, Nüsslein-Volhard C (1989) Determination of spatial domains of giant and Krüppel define middle body regions of the Drosophila embryo. De- zygotic gene expression in the Drosophila embryo by the affinity of binding sites for – the bicoid morphogen. Nature 340:363–367. velopment 111:611 621. 14. Hoch M, Schröder C, Seifert E, Jäckle H (1990) cis-acting control elements for Krüppel 35. Manu, et al. (2009) Canalization of gene expression in the Drosophila blastoderm by expression in the Drosophila embryo. EMBO J 9:2587–2595. gap gene cross regulation. PLoS Biol 7:e1000049. 15. Rivera-Pomar R, Lu X, Perrimon N, Taubert H, Jäckle H (1995) Activation of posterior 36. Manu, et al. (2009) Canalization of gene expression and domain shifts in the Dro- gap gene expression in the Drosophila blastoderm. Nature 376:253–256. sophila blastoderm by dynamical attractors. PLOS Comput Biol 5:e1000303. 16. Berman BP, et al. (2002) Exploiting transcription factor binding site clustering to 37. Yu D, Small S (2008) Precise registration of gene expression boundaries by a repressive identify cis-regulatory modules involved in pattern formation in the Drosophila ge- morphogen in Drosophila. Curr Biol 18:868–876. nome. Proc Natl Acad Sci USA 99:757–762. 38. Barolo S, Levine M (1997) Hairy mediates dominant repression in the Drosophila 17. Ochoa-Espinosa A, et al. (2005) The role of binding site cluster strength in Bicoid- embryo. EMBO J 16:2883–2891. dependent patterning in Drosophila. Proc Natl Acad Sci USA 102:4960–4965. 39. Winkler CJ, Ponce A, Courey AJ (2010) Groucho-mediated repression may result from 18. Margolis JS, et al. (1995) Posterior stripe expression of hunchback is driven from two a histone deacetylase-dependent increase in nucleosome density. PLoS ONE 5:e10166. – promoters by a common enhancer element. Development 121:3067 3077. 40. Li LM, Arnosti DN (2011) Long- and short-range transcriptional repressors induce 19. Struhl G (1989) Differing strategies for organizing anterior and posterior body pat- distinct chromatin states on repressed genes. Curr Biol 21:406–412. – tern in Drosophila embryos. Nature 338:741 744. 41. Papatsenko D (2007) ClusterDraw web server: A tool to identify and visualize clusters 20. Pankratz MJ, Busch M, Hoch M, Seifert E, Jäckle H (1992) Spatial control of the gap of binding motifs for transcription factors. Bioinformatics 23:1032–1034. gene knirps in the Drosophila embryo by posterior morphogen system. Science 255: 42. Markstein M, et al. (2004) A regulatory code for neurogenic gene expression in the 986–989. Drosophila embryo. Development 131:2387–2394. 21. Hartmann B, Reichert H, Walldorf U (2001) Interaction of gap genes in the Drosophila 43. Lee EC, et al. (2001) A highly efficient Escherichia coli-based chromosome engineering head: Tailless regulates expression of empty spiracles in early embryonic patterning and brain development. Mech Dev 109:161–172. system adapted for recombinogenic targeting and subcloning of BAC DNA. Genomics – 22. Häder T, et al. (2000) Receptor tyrosine kinase signaling regulates different modes of 73:56 65. fi Groucho-dependent control of Dorsal. Curr Biol 10:51–54. 44. Liu P, Jenkins NA, Copeland NG (2003) A highly ef cient recombineering-based 23. Venken KJ, He Y, Hoskins RA, Bellen HJ (2006) P[acman]: A BAC transgenic platform method for generating conditional knockout mutations. Genome Res 13:476–484. for targeted insertion of large DNA fragments in D. melanogaster. Science 314: 45. Kosman D, et al. (2004) Multiplex detection of RNA expression in Drosophila embryos. 1747–1751. Science 305:846.

6of6 | www.pnas.org/cgi/doi/10.1073/pnas.1109873108 Perry et al. Downloaded by guest on September 23, 2021