Large distances separate coregulated genes in livingDrosophila embryosTyler Heista, Takashi Fukayab,c, and Michael Levinea,d,1

aLewis-Sigler Institute for Integrative Genomics, Princeton University, Princeton, NJ 08544; bInstitute for Quantitative Biosciences, The University ofTokyo, 113-0032 Tokyo, Japan; cDepartment of Life Sciences, Graduate School of Arts and Sciences, The University of Tokyo, 113-0032 Tokyo, Japan;and dDepartment of Molecular Biology, Princeton University, Princeton, NJ 08544

Contributed by Michael Levine, June 11, 2019 (sent for review May 24, 2019; reviewed by Cornelis Murre and Christine Rushlow)

Transcriptional enhancers are short segments of DNA that switchgenes on and off in response to a variety of cellular signals. Manyenhancers map quite far from their target genes, on the order oftens or even hundreds of kilobases. There is extensive evidencethat remote enhancers are brought into proximity with theirtarget promoters via long-range looping interactions. However,the exact physical distances of these enhancer–promoter interac-tions remain uncertain. Here, we employ high-resolution imagingof living Drosophila embryos to visualize the distances separatinglinked genes that are coregulated by a shared enhancer. Cotransvectionassays (linked genes on separate homologs) suggest a surpris-ingly large distance during transcriptional activity: at least100–200 nm. Similar distances were observed when a shared en-hancer was placed into close proximity with linked reporter genesin cis. These observations are consistent with the occurrence of“transcription hubs,” whereby clusters (or condensates) of multi-ple RNA polymerase II complexes and associated cofactors are pe-riodically recruited to active promoters. The dynamics of thisprocess might be responsible for rapid fluctuations in the distancesseparating the transcription of coregulated reporter genes duringtransvection. We propose that enhancer-promoter communicationdepends on a combination of classical looping and linking models.

transcription | enhancers | transvection | Drosophila embryos |live-imaging

There is emerging evidence for transcription hubs, wherebygene activation is mediated by the recruitment of clusters or

condensates of multiple RNA polymerase II (Pol II) complexes(reviewed in ref. 1). Transient clusters of Pol II were detectedin living mammalian cells using high-resolution live imagingmethods (2). At least some of these clusters appear to be asso-ciated with foci of active transcription, visualized by the insertionof MS2 RNA stem loops into defined genes (3, 4). In vitro andin vivo assays suggest that the low complexity domains containedin many components of the preinitiation complex (PIC) mediateliquid-liquid condensation, including Pol II, TAF, and differentsubunits of the Mediator complex (5–7). Coordinate activationof linked reporter genes by a shared enhancer lends furthersupport for the occurrence of transcription hubs (8, 9).This hub model contrasts with the classical view of gene ex-

pression, which suggests the formation of a single PIC and re-cruitment of individual Pol II complexes at active promoters(10). An important implication of the hub hypothesis is thatdistal enhancers need not, and indeed cannot, come into closecontact with their target promoters due to molecular crowdingcaused by the collective size of multiple Pol II complexes andother components of the PIC (11, 12).Hi-C (chromosome conformation capture) contact maps have

been interpreted to suggest close contact of distal enhancers withtheir target promoters (13). However, recent high-resolutionimaging methods raise the possibility that distal enhancers canmap quite far during transcriptional activation. For example,activation of a synthetic lacZ reporter gene by endogenous even-skipped (eve) stripe enhancers is detected in living embryos over

apparent distances of 300–400 nm (14). Similar distances (∼300–350 nm) were observed between the Sox2 control region (SCR)enhancer and Sox2 promoter in mouse embryonic stem cells (15).High-resolution fixed tissue methods may be consistent with suchdistances in the Bithorax complex in Drosophila embryos (16).Here, we use a recently reported transvection assay and high-

resolution imaging of living Drosophila embryos to measure thedistances separating coexpressed reporter genes that are regu-lated by a single enhancer across homologous chromosomes (9).These studies reveal large distances of nascent transcripts asso-ciated with active alleles, on the order of 250–300 nm. Thesedistances include a relatively large standard deviation (SD)(∼120 nm). We present evidence that this variance is not strictlydue to technical or biological noise in our measurements, butinstead might arise from rapid fluctuations in the distancesseparating coordinately expressed alleles. We discuss the impli-cations of these measurements with the hub hypothesis and thedynamics of Pol II aggregation/condensation at active promoters.

Results and DiscussionAs a first step toward estimating the distance between a distalenhancer and its target promoter during gene activation, weperformed high-resolution measurements of coordinately activealleles during transvection (summarized in Fig. 1A). Briefly, thisassay employs a PP7 reporter gene with a closely linked upstreamsnail (sna) shadow enhancer, which mediates expression in thepresumptive mesoderm (17, 18). The other allele contains anMS2 reporter gene that lacks its own enhancer. Consequently,MS2 expression depends on pairing of the alleles to allow the sna

Significance

The prevailing model of gene activation is the looping of adistal enhancer to its target promoter, although it remainsuncertain whether enhancers must come into direct contact.Recent studies suggest that clusters of RNA polymerase II arerecruited to active genes, raising the possibility that enhancerscannot get too close to their target promoters due to molecularcrowding. Here, we use live imaging methods to visualizegenes that are simultaneously regulated by a single enhancerboth in cis and in trans. We present evidence that thesecoregulated genes are separated by a substantial distance—atleast 100–200 nm—during transcription. These observationschallenge the textbook view of enhancer function and raise thepossibility of an action-at-a-distance model for gene expression.

Author contributions: T.H. and M.L. designed research; T.H. and T.F. performed research;T.H. analyzed data; and T.H. and M.L. wrote the paper.

Reviewers: C.M., University of California San Diego; and C.R., New York University.

The authors declare no conflict of interest.

Published under the PNAS license.1To whom correspondence may be addressed. Email: msl2@princeton.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1908962116/-/DCSupplemental.

Published online July 8, 2019.

15062–15067 | PNAS | July 23, 2019 | vol. 116 | no. 30 www.pnas.org/cgi/doi/10.1073/pnas.1908962116

Dow

nloa

ded

by g

uest

on

Aug

ust 5

, 202

0

shadow enhancer on the PP7 allele to act in trans (9). To improvethe frequency of pairing and transvection, gypsy insulator DNAswere placed upstream of the 2 reporter genes.To address this question of distances separating enhancers and

target genes, we measured the center-to-center distances of ac-tive MS2 and PP7 foci in xy dimensions (representative nuclei are

shown in Fig. 1B). We focused on the xy axes due to technicaldifficulties in achieving similar resolution in the axial (z) di-mension. To compensate for this emphasis on xy resolution, weobserved a large number of transcription foci to ensure reliablecapture of the distances separating dynamically expressed al-leles. As a control, we analyzed a reporter gene that contains

A

B0 6.6 19.8 26.4 33.1 39.7 46.313.2 seconds

0 6.6 19.8 26.4 33.1 39.7 46.313.2 seconds

C

24xMS2 lacZ

eve Pr

enhancersna shadow

24xPP7 lacZ

eve Prgypsynu

cleu

s 1

nucl

eus

2

0.00

0.25

0.50

0.75

1.00

0 250 500 750 1000XY distance (nm)

Den

sity

MS2 His2AvPP7

Fig. 1. Large distances separate coregulated genes in trans. (A) A schematic of the transvection assay. On the top allele, there is a gyspy insulator, the snashadow enhancer, and a PP7-lacZ reporter gene. On the bottom allele, there is a gypsy insulator and an MS2-lacZ reporter gene. (B) Stills of 2 representativenuclei taken from maximum projected Airyscan-processed images. MS2 foci were visualized by MCP-GFP (green), PP7 foci were visualized by mCherry-PCP(red), and nuclei were visualized using His2Av-eBFP2 (blue). (Scale bar, 500 nm.) (C ) Distribution of xy distances separating MS2 and PP7 foci duringtransvection in nuclear cycle 14 (n = 2,441). Bin widths plotted were 10 nm.

Heist et al. PNAS | July 23, 2019 | vol. 116 | no. 30 | 15063

DEV

ELOPM

ENTA

LBIOLO

GY

Dow

nloa

ded

by g

uest

on

Aug

ust 5

, 202

0

A

24xMS2 yellow 24xPP7yellowenhancersna shadow

eve Pr eve Pr

enhancersna shadow

24xPP7 yellow

sna Pr

24xMS2yellow

sna PrB

C

D

0 9.9 29.7 39.6 49.5 59.4 69.319.8 seconds

0 10.3 30.8 41.2 51.5 61.7 72.020.5 seconds

E F

0

500

1000

1500

2000

Distal Proximal

Fluo

resc

ence

Inte

nsity

(AU

)

p = 1.2 x 10-46

dist

alpr

oxim

al

0.00

0.25

0.50

0.75

1.00

0 250 500 750 1000XY distance (nm)

Den

sity

proximal promoters

distal promoters

MS2 His2AvPP7

Fig. 2. Distal and proximal configurations of 2 cis-linked reporter genes yield unique distance distributions. (A and B) Schematics for 2 configurations with 2cis-linked reporter genes. The sna shadow enhancer is regulating MS2-yellow and PP7-yellow reporter genes with either distal eve promoters (A) or proximalsna promoters (B). The promoters are separated by either ∼16.5 kb (A) or ∼3.5 kb (B). (C and D) Stills of representative nuclei with either the distal (C) orproximal (D) configuration taken from maximum projected Airyscan-processed images.MS2 foci were visualized by MCP-GFP (green), PP7 foci were visualizedby mCherry-PCP (red), and nuclei were visualized using His2Av-eBFP2 (blue). (Scale bar, 500 nm.) (E) Distribution of xy distances separating MS2 and PP7 fociwith the distal (blue; n = 2,098) or proximal configuration in nuclear cycle 14 (red; n = 2,016). Bin widths plotted were 10 nm. These 2 distance distributionswere determined to be significantly different (P = 2.0 × 10−85, Mann–Whitney U test). (F) Boxplot showing distribution of fluorescence intensity of MS2 focibetween proximal (red; n = 832) or distal configuration (blue; n = 1,066). The box represents the lower (25%) and upper (75%) quantile and the solid linerepresents the median. Whiskers cover up to the 10th and 90th percentile of each distribution. P value was calculated using a Mann–Whitney U test.

15064 | www.pnas.org/cgi/doi/10.1073/pnas.1908962116 Heist et al.

Dow

nloa

ded

by g

uest

on

Aug

ust 5

, 202

0

interleaved MS2 and PP7 stem loops (19). These measurementssuggest a high degree of localization precision, incorporating bothtechnical noise and noise arising from the MS2/PP7 system, with32 ± 22 nm in the x axis and 26 ± 20 nm in y (SI Appendix, Fig. S1).During transvection, the 2 foci of transcription do not co-

incide, but instead display considerable spatial separation. Rapidfluctuations cause some variations in the distances separating thePP7 and MS2 nascent transcripts within individual nuclei duringthe acquisition window (e.g., Fig. 1B and Movie S1). We discussthese fluctuations in more detail below.Surprisingly large distances were found to separate the 2 al-

leles during transvection: 282 ± 124 nm (Fig. 1C), although theyare considerably smaller than those seen for MS2 and PP7 allelesthat are regulated by separate enhancers (SI Appendix, Fig. S2).This underscores the fact that coordinate transcription of PP7and MS2 by a shared enhancer depends on pairing of the 2 al-leles during transvection and the resultant increase in physicalproximity. Of course, these measurements represent only an ap-proximation of the distance separating the sna shadow enhancer

and MS2 target promoter since we are visualizing nascent tran-scripts. However, previous studies suggest that RNAs are tightlyassociated with the chromatin templates from which they are de-rived (20) and, therefore, could provide a reasonable proxy forvisualizing enhancer-promoter interactions across homologouschromosomes.The distances separating transvecting PP7 and MS2 alleles are

consistent with molecular crowding at active promoters due toaggregation of components of the general transcription ma-chinery (e.g., Pol II, Mediator). Further support for this viewstems from measurements of coordinately expressed PP7 andMS2 alleles linked in cis in close proximity on the same chro-mosome (Fig. 2). We measured the distances separating co-ordinately expressed MS2 and PP7 reporter genes organized indistal (Fig. 2A and Movie S2) or proximal (Fig. 2B and MovieS3) arrangements. The promoter regions of the reporter genesmap ∼16.5 kb away from one another in the distal arrangement.In this configuration, the distance between the transcription fociis an average of 202 ± 111 nm (Fig. 2 C and E), somewhat closer

A

13.7 48.1 61.9 103.1 seconds

B

0 13.2 33.1 52.9 seconds

C

59.5 86.0 92.6 125.6 seconds

D

52.9 59.5 72.7 125.6 seconds

0

200

400

600

800

0 50 100 150 200

XY D

ista

nce

(nm

)

0

200

400

600

800

0 50 100 150 200

XY D

ista

nce

(nm

)

0

200

400

600

800

0 50 100 150 200

XY D

ista

nce

(nm

)

0

200

400

600

800

0 50 100 150 200Time (seconds)

XY D

ista

nce

(nm

)

MS2 His2AvPP7

Fig. 3. Distance dynamics between coregulated genes during transvection. (A–D, Left) Stills of representative nuclei during transvection taken from max-imum projected Airyscan-processed images. MS2 foci were visualized by MCP-GFP (green), PP7 foci were visualized by mCherry-PCP (red), and nuclei werevisualized using His2Av-eBFP2 (blue). (Scale bar, 500 nm.) (A–D, Right) xy distance trajectories between MS2 and PP7 foci over time. Time points that areindicated by blue correspond to a still image shown at Left.

Heist et al. PNAS | July 23, 2019 | vol. 116 | no. 30 | 15065

DEV

ELOPM

ENTA

LBIOLO

GY

Dow

nloa

ded

by g

uest

on

Aug

ust 5

, 202

0

than that seen in the transvection assay. Surprisingly, the dis-tance separating the transcription foci is larger for the proximalarrangement, whereby the PP7 and MS2 promoters are sepa-rated by just ∼3.5 kb. The average distance that was measured is268 ± 119 nm, more similar to the distance seen in the trans-vection assay (Fig. 2 D and E).Why would the proximal arrangement exhibit a larger distance

than the distal arrangement? These distances correlate with thelevel of transcriptional activity (Fig. 2F). In the proximal orien-tation, the sna shadow enhancer is located immediately upstreamof both the MS2 and PP7 reporter genes, and the total fluores-cence intensity (burst size) is ∼33% higher than that observedbased on MS2 signal when the enhancer is positioned in thedistal orientation (8). We suggest that this difference in burst sizeis due to the different numbers of Pol II or associated factors(e.g., Mediator) that are recruited. More factors are recruitedwhen the enhancer is proximal, leading to a larger separation ofthe 2 reporter genes. In contrast, fewer factors might berecruited when the shared enhancer is located in the distal ori-entation. It is possible that this same phenomenon might occurbetween transgenic constructs inserted at different genomic lo-cations (SI Appendix, Fig. S3). An alternative possibility is thatthe greater distances separating the proximal vs. distal orienta-tions of the PP7 and MS2 reporters are due to divergent elon-gation of Pol II along reporter genes positioned in opposingdirections.As discussed earlier, there is a relatively large SD in our

measurements, ∼110–120 nm. This does not appear to be strictlydue to technical and biological noise (SI Appendix, Fig. S1), butrather, arises (at least in part) from rapid fluctuations in thephysical separation of the PP7 and MS2 reporter genes duringtheir coordinate expression (Fig. 3). We examined these fluctu-ations in >100 living nuclei and several examples are presentedin Fig. 3, e.g., the nucleus shown in Fig. 3D exhibits a full fluc-tuation cycle over an interval of ∼40 s. The timing and distancesof these fluctuations are variable, with distances of 50–200 nmand time intervals of ∼10–40 s (SI Appendix, Fig. S4).It is possible that these fluctuations are caused by the dynamic

recruitment of Pol II and associated cofactors. Perhaps Pol IIclusters of variable sizes are recruited to the hub. Large clusters(e.g., 10–20 Pol II complexes) lead to a larger separation (e.g.,∼200 nm) of the active reporter genes compared with smallclusters (5–6 Pol II complexes; ∼100 nm). As Pol II is releasedand undergoes elongation, the reporter genes may come intocloser proximity with one another. An alternate possibility is thatthese fluctuations might be a consequence of dynamic “kiss andrun” interactions, whereby an enhancer comes into sporadic andtransient contact with its target promoter to trigger transcription(4). However, we do not favor this view as we observed severalinstances of coregulated transcription foci that are separated bysubstantial distances (>200 nm) during the entirety of a fluctu-ation cycle (SI Appendix, Fig. S4).We believe that the transvection assay (Fig. 1) provides the

best opportunity for estimating the distance separating an en-hancer from its target promoter due to the tight linkage of thesna shadow enhancer with the PP7 reporter gene and theuncoupling of the MS2 target gene on the other homologouschromosome. As a result, the PP7 signals provide a reasonableestimate for the position of the enhancer. As discussed above,the mean distance separating PP7 and MS2 signals in thetransvection assay is ∼250–300 nm. However, this might over-estimate the actual distance between the PP7-linked enhancerfrom the MS2 target promoter, as we are measuring nascenttranscripts that are located downstream of the transcription startstate. If we conservatively apply our error in localization (SIAppendix, Fig. S1) to our measurements, the minimal distanceseparating the 2 alleles is then on the order of ∼200 nm. Further,with the fluctuations as described above, it is possible that

minimum distances of ∼100 nm separate the enhancer from itstarget promoter at transient periods during transcription, al-though distances of 200–300 nm (or more) are equally plausible.We suggest that even a distance of ∼100 nm is larger than what isexpected by traditional looping models.In summary, we have presented evidence that large distances

separate coordinately expressed reporter genes regulated by ashared enhancer. These distances are consistent with the hubmodel, whereby clusters or condensates of Pol II or associatedfactors are recruited to active foci (2–4, 21). Further support for thisview is the correlation between the distances of separation and thesize of transcriptional bursts (summarized in Fig. 4). Whether it isPol II or associated factors that are the major contributors ofmolecular crowding at active promoters is currently an area of ac-tive study (2–4, 21). Nonetheless, the magnitude of the distancesdocumented in this study is consistent with several recent studiesthat have employed various measurement methods, in both fixedand living tissues (14–16). The picture that is emerging from thesestudies is that enhancer-promoter communication is mediated by acombination of the classical looping and linking models. Once en-hancers come into proximity, they influence transcription via manyprotein–protein interactions within a hub across the distance of theenhancer to its promoter.

Materials and MethodsFly Strains. The fly strains gypsy-sna shadow enhancer-evePr-PP7-lacZ (9),gypsy-sna shadow enhancer-evePr-MS2-lacZ (9), gypsy-evePr-MS2-lacZ (9),

A

enhancer

MS2

PP7

~280nm

~200nm

~270nm

recn

ahne

MS2

PP7

B

C

r ecn

ahne

MS2

PP7

Fig. 4. Molecular crowding as the source for the distances separatingcoregulated genes in different configurations. (A–C) In each assay, we proposethat clustering and aggregation of Pol II (purple), transcription factors(orange), and other components of the general transcription machinery (e.g.,PIC components, Mediator; yellow) leads to the separation seen betweencoregulated reporter genes. With the cis-linked assays, recruitment of fewersuch molecules (e.g., Pol II, Mediator) might explain the smaller distanceobserved between reporter genes in the distal configuration (B) comparedwith the proximal configuration (C).

15066 | www.pnas.org/cgi/doi/10.1073/pnas.1908962116 Heist et al.

Dow

nloa

ded

by g

uest

on

Aug

ust 5

, 202

0

evePr-MS2-yellow-sna shadow enhancer-yellow-PP7-evePr (8), snaPr-MS2-yellow-sna primary enhancer-yellow-PP7-snaPr (8), and hb proximalenhancer-hb P2 promoter-12×(PP7-MS2)-lacZ (19) used in this study werepreviously published and have all been integrated into the third chromo-some (VK00033). The pbphi-snaPr-MS2-yellow-sna primary enhancer-yellow-PP7-snaPr plasmid (8) was also injected and integrated into the secondchromosome (VK00002) by BestGene.

Cloning and Transgenesis. The pbphi-yellow-MS2-snaPr-linker-sna shadow-linker-snaPr-PP7-yellow plasmid was constructed as previously described (8),except the insertion of the DNA fragment containing linker-snaPr-MS2-yellow in the opposite orientation was selected after sequencing. Injection wasdone as previously described (22) into the third chromosome (VK00033).

Live Imaging. For the assays containing both MS2 and PP7 transgenes on thesame allele, MCP-GFP, mCherry-PCP, His2Av-eBFP2 homozygous virgins (9)were mated with homozygous males containing the transgenic construct.For the assays containing MS2 and PP7 transgenes on separate alleles, MCP-GFP, mCherry-PCP, His2Av-eBFP2 homozygous virgins were first mated withhomozygous males carrying the PP7-containing transgenic construct. Re-sultant trans-heterozygote virgins were obtained and mated with homozy-gous males carrying the MS2-containing transgenic construct. For all crosses,resulting embryos were treated in preparation for imaging as previouslydescribed (9). All imaging was done at room temperature using a Zeiss LSM880 with the Fast Airyscan module and using a Plan-Apochromat 63×/1.4N.A. oil immersion objective. The 405-nm, 488-nm, and 561-nm lasers wereused across all imaging. For the transvection assay, 30 z-slices of 0.17 umwere acquired per time point. For all other experiments, either 45 or 60 z-slicesof 0.17 um were acquired per time point. For the experiments examiningthe 2 cis-linked reporter genes at different integration sites (SI Appendix,Fig. S3), 0.7× Nyquist xy sampling was used. For all other experiments, 1.8×Nyquist xy sampling was used. For assays in which we compared fluores-cence intensities across constructs (Fig. 2), laser power was kept constant.Images were captured in 8 bits.

Image Analysis. All images were first deconvolved and processed in Zen Black,version 2.3 (Zeiss). Processed images were then moved to Imaris, version 9.2.1(Bitplane). Objects were fit to MS2 and PP7 foci using the Surfaces tool andtracked across time in the processed movie. For each time point in eachnucleus, the centers of both foci were recorded. Time points in which a singleMS2 or PP7 focus was split into 2 distinct foci, likely due to sister chromatidseparation, were excluded from further analysis by manual curation. Four-color 200-nm diameter Tetraspeck beads (Thermo Fisher) were imaged asdescribed above, and the centers of each bead signal upon 488-nm and 561-nmexcitationwere calculated. The optical offsets in these channels were usedto fit a second-order polynomial transform in MATLAB R2016b (Mathworks)using the cp2tform function (23). This mapping was used to correct the offsetbetween the 2 color channels in the experimental images. For Fig. 2F,fluorescence intensity was determined as the highest intensity pixel found ineach MS2 surface per time point and nucleus (this is due to the higherphotostability of GFP vs. mCherry).

Statistical Analysis. All plots and statistical tests were done in R, version 3.5.1.A 2-sided Mann–Whitney U test was used to calculate statistical significanceof the difference between 2 distributions of fluorescence intensity (Fig. 2F)and between 2 distributions of distance (Fig. 2E and SI Appendix, Fig. S3D).

ACKNOWLEDGMENTS. We thank Evangelos Gatzogiannis for instructionand assistance with imaging, the Shvartsman laboratory for assistance withImaris, as well as Paul Schedl, Stas Shvartsman, Benjamin Bratton, andmembers of the M.L. laboratory for helpful discussions. T.F. was supportedby the Grants-in-Aid for Leading Initiative for Excellent Young Researchersfrom the Ministry of Education, Culture, Sports, Science and Technology inJapan, and the Grants-in-Aid for Research Activity Start-up JP18H06040 andthe Grants-in-Aid for Scientific Research B JP19H03154 from the JapanSociety for the Promotion of Science, and research grants from the MolecularBiology Society Japan, the Mochida Memorial Foundation for Medical andPharmaceutical Research, the Inamori Foundation, and the Nakajima Founda-tion. This study was funded by NIH Grant GM118147 (to M.L.).

1. E. E. M. Furlong, M. Levine, Developmental enhancers and chromosome topology.Science 361, 1341–1345 (2018).

2. I. I. Cisse et al., Real-time dynamics of RNA polymerase II clustering in live human cells.Science 341, 664–667 (2013).

3. W. K. Cho et al., RNA Polymerase II cluster dynamics predict mRNA output in livingcells. eLife 5, e13617 (2016).

4. W. K. Cho et al., Mediator and RNA polymerase II clusters associate in transcription-dependent condensates. Science 361, 412–415 (2018).

5. A. Boija et al., Transcription factors activate genes through the phase-separationcapacity of their activation domains. Cell 175, 1842–1855.e16 (2018).

6. B. R. Sabari et al., Coactivator condensation at super-enhancers links phase separationand gene control. Science 361, eaar3958 (2018).

7. Y. Shin et al., Liquid nuclear condensates mechanically sense and restructure thegenome. Cell 175, 1481–1491.e13 (2018). Erratum in: Cell 176, 1518 (2019).

8. T. Fukaya, B. Lim, M. Levine, Enhancer control of transcriptional bursting. Cell 166,358–368 (2016).

9. B. Lim, T. Heist, M. Levine, T. Fukaya, Visualization of transvection in living Drosophilaembryos. Mol. Cell 70, 287–296.e6 (2018).

10. R. G. Roeder, The complexities of eukaryotic transcription initiation: Regulation ofpreinitiation complex assembly. Trends Biochem. Sci. 16, 402–408 (1991).

11. P. Cramer, D. A. Bushnell, R. D. Kornberg, Structural basis of transcription: RNA po-lymerase II at 2.8 angstrom resolution. Science 292, 1863–1876 (2001).

12. S. Schilbach et al., Structures of transcription pre-initiation complex with TFIIH andMediator. Nature 551, 204–209 (2017).

13. B. Mifsud et al., Mapping long-range promoter contacts in human cells with high-resolution capture Hi-C. Nat. Genet. 47, 598–606 (2015).

14. H. Chen et al., Dynamic interplay between enhancer-promoter topology and geneactivity. Nat. Genet. 50, 1296–1303 (2018).

15. J. M. Alexander et al., Live-cell imaging reveals enhancer-dependent Sox2 transcrip-tion in the absence of enhancer proximity. elife 8, e41769 (2019).

16. L. J. Mateo et al., Visualizing DNA folding and RNA in embryos at single-cell resolu-tion. Nature 568, 49–54 (2019).

17. M. W. Perry, A. N. Boettiger, J. P. Bothma, M. Levine, Shadow enhancers foster ro-bustness of Drosophila gastrulation. Curr. Biol. 20, 1562–1567 (2010).

18. L. Dunipace, A. Ozdemir, A. Stathopoulos, Complex interactions between cis-regulatory modules in native conformation are critical for Drosophila snail expres-sion. Development 138, 4075–4084 (2011).

19. T. Fukaya, B. Lim, M. Levine, Rapid rates of Pol II elongation in the Drosophila em-bryo. Curr. Biol. 27, 1387–1391 (2017).

20. J. D. Larkin, A. Papantonis, P. R. Cook, D. Marenduzzo, Space exploration by thepromoter of a long human gene during one transcription cycle. Nucleic Acids Res. 41,2216–2227 (2013).

21. J. Li et al., Single-molecule nanoscopy elucidates RNA polymerase II transcription atsingle genes in live cells. Cell, 10.1016/j.cell.2019.05.029 (2019).

22. L. Ringrose, Transgenesis in Drosophila melanogaster. Methods Mol. Biol. 561, 3–19(2009).

23. M. Erdelyi et al., Correcting chromatic offset in multicolor super-resolution localiza-tion microscopy. Opt. Express 21, 10978–10988 (2013).

Heist et al. PNAS | July 23, 2019 | vol. 116 | no. 30 | 15067

DEV

ELOPM

ENTA

LBIOLO

GY

Dow

nloa

ded

by g

uest

on

Aug

ust 5

, 202

0