Vol. 7, 1393-1401, October 1996 Cell Growth & Differentiation 1393

Transcriptional Specificity of the Pluripotent EmbryonicStem Cefl1

Christina A. Scherer,2’3 Jin Chen, Abudi Nachabeh,Nancy Hopkins, and H. Earl Rule?

Center for Cancer Research and Department of Biology, MassachusettsInstitute of Technology, Cambridge, Massachusetts 02139 [C. A. S.,N. H.], and Department of Microbiology and Immunology, VanderbiltUniversity School of Medicine, Nashville, Tennessee 37232-2363[J. C., A. N., H. E. R.]

AbstractThe specificity of gene expression in embryonic stem(ES) cells was analyzed both under in vitro cultureconditions and during early embryogenesis. ES cellswere infected with U3figeo, a U3 gene trap retrovirusthat contains coding sequences for a �-galactosidase-neomycin phosphotransferase hybrid protein.Integrated proviruses, which disrupted expressedcellular genes, were selected in the presence of G418.ES clones expressing regulated I3geo fusion geneswere identified by changes in 5-bromo-4-chloro-3-indolyl-�3-D-galactopyranoside staining after in vitrodifferentiation. Thirty-one of 191 clones tested (16%)exhibited regulated expression of �Jgeo protein. Sevengenes disrupted by U3�geo were passed into thegermline, and expression of the IJgeo fusion genes wasanalyzed in vivo, including inserts disrupting the Eckand REX-I genes. In each case, genes trapped incultured ES cells were expressed in the inner cell massof preimplantation embryos, and changes in IacZexpression during in vitro differentiation were alsoobserved during early development. Thus, cultured EScells maintain, to a considerable extent, thetranscriptional specificity of the pluripotent cells of thepreimplantation embryo. As a consequence, in vitroscreens utilizing gene traps provide a rapid andaccurate means to identify and disrupt developmentallyregulated genes.

Received 5/10/96; revised 7/16/96; accepted 7/29/96.The costs of publication of this article were defrayed in part by thepayment of page charges. This article must therefore be hereby markedadvertisement in accordance with 1 8 U.S.C. Section 1 734 solely to mdi-cate this fact.1 This work was supported by NIH Grant RO1GM84688 (to H. E. A.).C. A. S. was supported by National Research Service Award TrainingGrant GM07187 and by the Massachusetts Institute of Technology De-partment of Biology. J. C. was supported by American Cancer SocietyGrant PF-3689 and National Institute of General Medicine GrantF32GM1 7003.2 Present address: Department of Microbiology, University of Washington,Seattle, WA 98195.3 C. A. S. and J. C. contributed equally to this study.4 To whom requests for reprints should be addressed, at Department ofMicrobiology and Immunology, Room AA5206 MCN, Vanderbilt UniversitySchool of Medicine, 1 161 21 st Avenue South, Nashville, TN 37232-2363.Phone: (615) 343-1379; Fax (615) 343-7392.

IntroductionEvents during early mammalian development have been dif-ficult to study because of the inaccessibility of the implanted

embryo. Consequently, there has been considerable interest

in using mouse ES5 cells as a model to study early develop-ment in vitro. ES cells most closely resemble the pluripotentcells in the 1CM of blastula-stage embryos, which give rise tothe embryo proper and to several extraembryonic tissues (i,

2). Most significantly, ES cells can be cultured for extendedperiods without compromising their ability to form all embry-

onic cell types when reinserted into the blastocyst (3-6). Inthe absence of factors that inhibit differentiation, ES cells

form structures known as embryoid bodies, which contain anumber of cell types, including endoderm, ectoderm, skele-

tal, cardiac and smooth muscle, neuron, and nucleated redblood cells (1 , 2, 7). By manipulating the extracellular envi-ronment, patterns of differentiation can be made to favorindividual cell types (8-14).

In principle, cultured ES cells may be used to study themolecular determinants of the pluripotent state and the proc-ess by which early stem cells become committed to specificprograms of cell differentiation. Conceptually, both problems

concern the process by which external stimuli remodel thetranscriptional program of the cell. Therefore, candidategenes important for early development are expected to in-dude developmentally regulated genes and the transcription

factors that regulate their expression. As a first step, genesregulated during ES cell differentiation can be isolated eitherby differential cDNA cloning (1 5) or by gene-trapping strat-

egies (1 6-i 8). The latter approach directly generates mutant

alleles to facilitate studying gene functions in vivo.

The usefulness of ES cells as a model for early mammaliandevelopment will depend greatly on whether the transcrip-tional regulation upon ES cell differentiation in vitro reflectsgene regulation during development in vivo. Concerns havebeen raised as to whether many genes expressed in ES cellsare simply a consequence of cell culture conditions. In ad-dition, the genome of cultured ES cells, like their counter-parts in the 1CM, is globally hypomethylated (19, 20). Thisraises the possibility that many genes may be expressedgratuitously because of widespread activation of thegenome.

There have been conflicting reports concerning the spec-ificity of gene expression in cultured ES cells. Some genes

are regulated in the same manner both in vitro and in vivo (15,

21-26). However, other genes expressed in cultured ES cellsare not expressed in the 1CM (15, 27). For example, the

5 The abbreviations used are: ES, embryonic stem; 1CM, inner cell mass;LTR, long terminal repeat; nt, nucleotide; AT, read through; pc, postco-itum; RACE, rapid amplification of cDNA ends; X-GaI, 5-bromo-4-chloro-3-indolyl-f3-D-galactopyranoside.

1394 Transcriptional Specificity in Mouse Stem Cells

Lectl 4 and calcylin genes are expressed in ES cells and

repressed during differentiation but are not expressed in the1CM, as assessed by either in situ hybridization or immuno-staining (15, 27).

A large-scale gene-trap screen for developmentally regu-lated genes has been reported (28). The screen was basedon gene expression in 8.5-day chimeric mouse embryos.Because none of the cell-IacZ fusion genes was transmittedto the germline, expression in blastocyst and gastrula-stageembryos was not examined. Although this study identifiedgenes with restricted expression patterns, the correlationbetween the transcriptional regulation during ES cell differ-

entiation in vitro and gene regulation during early develop-ment in vivo was not addressed.

Because of these issues, we have compared the transcrip-tional specificity of the pluripotent stem cell with that of earlyembryos. For this analysis, a gene trap retrovirus (U3pgeo)was used to create random insertion of a IacZ-neo (j3geo)

reporter gene into genes expressed in ES cells. To avoidconfusion, the term “fusion gene” will always refer to a �3geosequence that is transcribed from a flanking cellular pro-motor and not to the �-galactosidase-Neo fusion protein.The expression of each fusion gene was monitored afterdifferentiation of clones in vitro. Selected fusion genes wereintroduced into the germline, and patterns of �-galactosid-ase gene expression in pro- and postimplantation embryoswere compared to those observed in vitro. We have analyzedthe expression of 191 fusion genes in cultured stem cellsbefore and after differentiation. Seven regulated fusion genesin this study and three fusion genes identified previously (29)were analyzed after germline transmission. In all 10 cases,expression of the fusion gene in ES cells accurately pro-dicted regulated transgene expression dunng early develop-ment. These studies show that the regulation of gene ex-pression in cultured pluripotent stem cells typically reflectstranscriptional control during early development.

Results

Generation of the U3I3geo Gene Trap Retrovirus.pU3f3geo was derived from pGgTKNeoU3LacZen(-) (30) byreplacing !acZ sequences with �3geo sequences (1 7). Celllines producing ecotropic U3f3geo retroviruses were goner-ated by transfecting pU3�3geo into ‘P2 cells (31). The titer ofthe producer line used as the source of virus in these exper-iments (‘I’2-�3geo2) was 2 x 10� G41 8R colony-forming unitsper ml per 106 producer cells as assayed on NIH 3T3 cells.Because the vector contains no independent selectablemarker, only those clones that contain proviruses insertedinto expressed genes will survive G41 8 selection. On thebasis of our experience with other gene trap vectors (30, 32),we estimate that fewer than 1 in 200 proviral integrationsallow U3 gene expression. Therefore, virus production by‘I’2-�3geo2 appears to be comparable to other Moloney mu-rine leukemia virus-based vectors, suggesting that insertionof f3geo sequences into U3 does not significantly affect viralinfectivity.

The structures of integrated proviruses in both NIH 3T3and ES cells were analyzed by Southern blot analysis. Al-though no mutant proviruses were observed in 3T3 clones,

approximately 80% of the ES clones contained deletionswithin the provirus. Southern blot analysis of these deletionsindicated that most of the ES cell clones contained a singleviral U3j3geo LTR (data not shown). For example, genomicsequences flanking the integration site in both J3A3 cells andwild-type cells were cloned into A DashlI and A Fixll (Strat-agene), respectively. Comparison of the virus-cell junction

sequences with the unoccupied site revealed a single LTRinsert, loss of two nucleotides from the ends of the U3 andU5 regions, and duplication of four nts of cellular DNA, as isnormal for Moloney murine leukemia virus proviruses (33).This is consistent with our previous results (1 8, 29) indicatingthat deletions within the provirus do not alter the remaining

viral LTR and flanking cellular sequences.X-Gal Staining Reflects the Level of Reporter Gene

Transcription. Clones expressing j3geo fusion genes typi-cally produce �-gaIactosidase-neomycin phosphotrans-ferase fusion protein. In fact, 65% of the ES cell clonesexhibited varying degrees of X-GaI staining, ranging fromvery light blue to dark blue, whereas 35% of the clones didnot appear to stain with X-Gal. To test whether the U3�geoprovirus functions as a gene trap in white as well as blue celllines, RNase protection assays and Northern blot analysiswere performed (Fig. 1).

Transcripts that result in the expression of U3pgeo aredepicted in Fig. 1A. Integration of the provirus into tran-scribed loci is characterized by the expression of a fusiontranscript originating in the flanking cellular DNA. This fusiontranscript can be distinguished from transcripts extendingthrough the 3’ LTR by RNase protection analysis. Hybridiza-tion to a 689-nt riboprobe derived from a region containingportions of env, U3, and IacZ will protect a 501 -nt fragmentif the transcript initiates in flanking cellular DNA (designated5’ AT) and a 643-nt fragment if the transcript extendsthrough the 3’ LTR (3’ RT). Indeed, RNA from all G418Rclones protected a 501 -nt fragment (Fig. 1B, Lanes 1-10; afragment is seen in Lane 3 upon prolonged exposure). Thisband is not seen in control tRNA and uninfected ES cell lanes(Fig. 1B, Lanes 1 1 and 12). The amount of 5’ RT detectedwas consistently higher in RNA from blue clones (Lanes

4-10) than RNA from white clones (Fig. 1B, Lanes 1-3),indicating that less of the fusion transcript was expressed inwhite clones. For comparison, RNA from all clones protectedsimilar levels of a 280-nt fragment derived from transcripts ofthe L32 large ribosomal subunit. The level of fusion transcriptexpressed in all clones was much lower than the L32 control,which represents a highly expressed housekeeping gene[comparing the intensity of the L32 control (Fig. 1B, Lane 12)

with the intensity of the 501 -nt bands (Fig. 1B, Lanes 1-10),

which were exposed for comparable times]. RNA fromclones containing full-length proviruses also protected smallamounts of a 643-nt fragment (3’ RT, Fig. 1B, Lanes 1, 2, 6,

and 8). Because the internal viral promoter is inactive inundifferentiated stem cells (4), the presence of this bandusually indicates transcripts that failed to become polyade-nylated in the 5’ LTR and extended through the 3’ LTR.

Northern blot hybridization confirmed the presence offJgeo transcripts in G418R clones (Fig. 1C; Lanes 1-9 corre-spond to Lanes 2-10 in Fig. 1B). The observed transcripts

A.

-

Cell Growth & Differentiation 1395

B.

5’RT 43kb 3’RT�>. �r � � 0’

iiiHHIuiII 13900 1II_�:� I � 1�FI)hhhhhhh1

U3 RU5 SupF U3 RU5

1 2 3 4 5 6 7 8 9 10 11 12

- . � - �. � .- -, .� .� -� .. - ._#{149}

C.

1234567 8 9 10

a

�4- P robe

-‘#{248}----3’RT

‘4- 5’ RI

-*-L32

. ,..‘. : .4- 28S

Fig. 1 . Analysis of cell-provirus fusion transcripts in clones infected with U3f3geoSupF. Total RNA was isolated from ES cells and analyzed in RNase protectionassays and by Northem blot hybridization. A, transcripts expected in NeoR clones. Solid arrow, the major fusion transcript seen by Northern blot hybridization.Small amounts of fusion transcript are not polyadenylated in the 5’ LTR and extend through the 3’ LTR (stippled arrow). Hybridization of RNA to a 689-nt lacZriboprobe protects a 501 -nt fragment (5’ R7) from transcripts initiating in the cellular DNA upstream of the provirus and a 643-nt fragment (3’ Ri) in transcriptsextending through the 3’ LTR. B, RNase protection analysis of fusion transcripts. Thirty �g of total RNA were hybridized simultaneously to a 689-nt nboprobecomplementary to sequences in env, U3, and LacZ and a 280-nt nboprobe complementary to the nbosomal L32 transcript (internal contro�. RNA from all NeoRclones protects a 501 -ntfragment indicative oftranscripts initiating in flanking cellular DNA. Lanes 1-3, RNAfrorn white clones; Lanes 4-10, RNA from blue clones;Lane 1 1, a yeast tRNA control; Lane 12, an uninfected ES cell control. Lane 12 has been overexposed to show that no lacZ transcnpts are detected. C, Northernblot hybridization of 10 j�g RNA from ES cell clones hybridized to a Nec probe. Lanes 1-9 correspond to Lanes 2-10 in B. Cellular-proviral fusion transcripts areonly evident in blue clones (Lanes 3, 5, 6, 8, and 9). Fusion transcripts from white and light blue clones (Lanes 1, 2, 4, and 7) were not detected by Northern blotanalysis. Lane 10 contains RNA from uninfected ES cells.

vary in size from 4.5 to 6 kb, which is consistent with the

expected size of transcripts that initiate in flanking cellular

DNA and terminate in the 5’ LTR. Because Northern blot

analysis is less sensitive than RNase protection, fusion tran-

scripts were detected in most blue clones (Fig. 1 C, Lanes 3,5, 6, 8, and 9) but not in white (Fig. 1 C, Lanes 1 and 2) or light

blue (Fig. 1C, Lanes 4 and 7) clones. These Northern and

RNase protection assays indicate that the intensity of X-Gal

staining provides a reliable indication of reporter gene ex-

pression. In addition, they show that a difference in the

sensitivity of the assays exists for neomycin phosphotrans-

ferase and j3-galactosidase activity in the pgeo fusion pro-

tein. Neomycin provides a very efficient selection for gene

trapping, in that there is no obvious bias against activation of

U3I3geo by weakly expressed genes.

Selection of Clones Exhibiting Regulated Fusion GeneExpression. To screen for changes in reporter gene expres-

sion upon in vitro differentiation, embryoid bodies were pro-

duced from 191 different G418R clones and stained with

X-Gal. For the purposes of this study, only clones that

,� �._J�_

. ..‘ �

,,c � � ‘ � - ,fr � . ,‘ -,� . .� .

-�. . , .�‘-. ... ‘ .�

,‘ ., ‘;.� 1�;.� -. .�-- ‘�

,.. .‘

�- , A

.-�, --- � .. “.4

� ...- ..4

� - --‘ -

� �.. ,.#. � ,-� F’

- . ,,A � _ .

. .-. .. ... � �. ..,. ... - � . , “ �. ‘

‘?�‘�“� � #{149} � � �.“ . ,. � �

. �.. - � � � “� .:

. - . � �,

. � .,‘:-�,_‘.

. - -�--. -. .,�.. �r.- �..-.- ‘�.t .‘.

�‘6� � F

H

1396 Transcriptional Specificity in Mouse Stern Cells

‘S

��#{149}‘ai #{149}

A

� I � �

..:�Y�.�:#{149}”:.-y�‘: � -�::i: -: � :� *:.. , .

� .‘q

�-

� E

� ‘�i ‘ . � .‘ � ;:i# .\..�,. .� �

�1-,,”,. � � f � �.‘, .. ‘I.,.’, 71/�� I � � � -

.. - ‘ ‘1 � � ‘ � � � � ‘

. �1 � ,;.# .,,, /

.�.. ..- 1 ,# � .‘ .:;�‘‘�

.�.. ‘ -.- . .:

.. ,..� -,,�- -

�Fig. 2. Expression of U3(3geo fusion genes in un-differentiated and differentiated ES cells. Undiffer-entiated ES cells (A, C, E, and G) and ernbryoidbodies from the same clones (B, D, F, and H) werefixed and stained with X-GaI. The clones show thefollowing staining patterns: A and B, constitutiveexpression (blue to blue); C and D, little to no stain-ing in either ES cells or embryold bodies (white towhite); E and F, repression upon differentiation (blueto white); and G and H, induction upon differentiation(white to blue).

.� � � - . � � �-5. PT

- � -��- :� �a . -- � �-L32

Fig. 3. RNase protection analysis of RNA from undifferentiated and dif-ferentiated ES cells. Total RNA was obtained from undifferentiated EScells (U) and embryoid bodies (D), and 30 .tg were used for RNaseprotection analysis as described in Fig. 1 . X-Gal staining in clone 7.4.2was induced upon differentiation, whereas staining in clones B2-3, 2.4,J1D4, J3A3, and J5C1 was reduced.

Cell Growth & Differentiation 1397

6 � A. Scherer, J. Chen, M. Pawlak, and H. E. Ruley, unpublished results.

7.4.2 J2B4 B2-3 2 4 Ji D4 J3A3 J5C1U D U D Li D U D U D U 0 U D

showed obvious differences in expression were analyzed

further. A total of 3i clones (i6%) exhibited regulated ex-

pression of pgeo upon in vitro differentiation; expression of

24 fusion genes (i 9% of blue clones) was repressed (blue to

white), whereas 7 (1 0% of white clones) showed increased

expression (white to blue). Typical staining patterns ob-

served in undifferentiated and differentiated clones are

shown in Fig. 2. Fig. 2, A-D, shows clones that did not exhibit

regulated expression, i.e., blue to blue (Fig. 2, A and B) and

white to white (Fig. 2, C and D), and Fig. 2, E-H, shows

examples of a blue-to-white clone (Fig. 2, E and F) and a

white/light blue-to-blue clone (Fig. 2, G and H).

To test whether differences in X-Gal staining reflected

changes in the amount of fusion transcript produced, RNase

protection analysis was performed on total RNA isolated

from undifferentiated cells and embryoid bodies. Fig. 3

shows data from seven clones that were later used to pro-

duce germline chimeras. Lanes labeled U contain RNA from

undifferentiated cells, and lanes labeled D contain RNA from

embryoid bodies. There is a significant decrease in the

amount offusion transcript (5’ Ri) in all clones in which f3geo

expression was repressed upon differentiation (B2-3, 2.4,

Ji D4, J3A3, and J5Ci), whereas levels of the L32 transcript

in each clone were unchanged. Clone 7.4.2, which showed

an increase in X-Gal staining upon differentiation in vitro,

exhibited a corresponding increase in the amount of 5’ RT

RNA. Clone J2B4, on the other hand, which appeared to be

up-regulated in vitro, exhibited no change in the amount of

fusion transcript. An increase in the amount of 3’ RT RNA is

also seen in clone 7.4.2, which contains a full-length provi-

rus. This increase is most likely due to either an increase of

transcripts extending through the 3’ LTR or activation of the

viral promoter upon differentiation. In summary, six out of

seven clones exhibited changes in the amount of fusion

transcript expressed that correlate with the observed X-GaI

staining patterns. The seventh clone did not show any sig-

nificant changes in the amount of fusion transcript

expressed.

Differential X-GaI Staining in Vitro Typically PredictsFusion Gene Expression in Transgenic Mice. Seven fu-

sion genes exhibiting regulated expression in vitro were

transmitted to the germline. Three additional lines, 3A8 (reg-

ulated expression), X5 (constitutive expression), and 2E7

(constitutive expression), expressing fusion genes that were

isolated by flow cytometry (29), were also analyzed. Het-

erozygous males were mated to C57B1J6 females, and n-ga-

�.-3. RT lactosidase expression was assayed in blastocysts (day 3.5

pc) and implanted embryos (days 6.5-12.5 pc). X-Gal stain-

ing of heterozygous embryos showed that the changes in

expression observed upon in vitro differentiation of ES cells

predicted those seen in postimplantation embryos (summa-

rized in Table 1 ). Three categories of regulated gene expres-

sion were observed in blastocysts and postimplantation em-

bryos (Fig. 4): (i) blue to restricted blue; (ii) white (or very weak

staining) to blue; and (iii) blue to white. The first and third

categories include the five clones that exhibited restricted

expression of the fusion transcript upon in vitro differentia-

tion. Category i consists of genes that are expressed in

blastocysts but show restricted staining patterns later in

development (Fig. 4, A and B). Staining in Ji D4 embryos

becomes restricted to the extraembryonic tissue at day 7.5

pc (Fig. 4A). Staining in J3A3 embryos becomes restricted to

a small group of cells in the node (Fig. 4B) and the developing

hindbrain (not shown) in day-8.0 pc embryos. Category ii

includes clones 7.4.2 and J2B4, which increase expression

upon differentiation in vitro. Clone 7.4.2 did not show staining

at the blastocyst stage but exhibited widespread staining in

postimplantation embryos, which is more intense in the brain

and neural tube in 8.5-day pc embryos (Fig. 4B). Clone J2B4,

which is expressed at a low level in blastocysts, is expressed

constitutively in postimplantation embryos (Fig. 4E). Cate-

gory iii includes B2-3, 2.4, and J5C1 , which stained blue or

light blue as blastocysts but did not stain embryos ages

6.5-1 0.5 days pc. An example of one of these clones is

shown in Fig. 4D, in which stained blastocysts are shown on

the left and day-8.5 pc embryos are shown on the right. X5

and 2E7, two constitutively expressed clones during ES cell

differentiation, are expressed constitutively during develop-

ment (29).

Identification of Disrupted Genes. Upstream cellular se-

quences flanking all seven U3f3geo proviruses were isolated

rapidly either by inverse PCR or 5’ RACE. In some cases, the

flanking cellular sequences obtained by PCR were sufficient

to identify the disrupted genes. For example, sequence anal-

ysis of the 5’ RACE product from the Ji D4 clone revealed

that the provirus had integrated into the REX-i gene (34).

Two of the clones that either resulted in embryonic lethality

in homozygotes or showed restricted expression pattern

during early embryogenesis were analyzed additionally. In

the case of clone 7.4.2, the 5’ RACE product was used to

isolate cDNAs from an 8.5-day embryonic library. The pre-

dicted amino acid sequence of the 7.4.2 open reading frame

is nearly identical to that of a rat protein arginine methyl

transferase.6 Mice homozygous for the 7.4.2 provirus exhibit

an overt phenotype (embryonic death). For clone J3A3, the

flanking probes were used to isolate genomic clones. Anal-

ysis of genomic clones revealed that the provirus had inte-

grated into the Eck receptor tyrosine kinase gene (33, 35, 36).

A detailed analysis of J3A3 (Eck) has been reported (33), and

detailed analysis of other clones will be published elsewhere.

1398 Transcriptional Specificity in Mouse Stem Cells

7 J. Chen and C. A. Scherer, unpublished results.

8 The present study and G. Hicks, E-G. Shi, M. Roshon, D. Williamson,

and H. E. Ruley, manuscript in preparation.

Table 1 Summary of germline clones

Clone In vitro regulation Blastocysts Day-8.5 embryos

7.4.2 White to blue White Diffused, high level in brain and eye

J2B4 Ught blue to blue Light blue Widespread blueB2-3 Blue to white Blue White2.4 Blue to white Blue WhiteJi D4 (Aexi) Blue to white Blue Restricted, extraembryonic ectodermJ3A3 (Eck) Blue to white Blue Restricted, node and hindbrainJ5C1 Blue to white Blue White3A8 Blue to white Blue Restricted, neural foldX5 Constitutive blue Blue Blue2E7 Constitutive blue Blue Blue

In all cases analyzed, the patterns of X-GaI staining in thetransgenic embryos was nearly identical to the tissue distri-bution of normal gene expression as assessed by in situ

hybridization or immunohistochemistry (26, 35, 36).�

DiscussionIn principle, the ability to identify and mutate developmentallyregulated genes in vitro could facilitate studies of specificdevelopmental processes. However, it has not been estab-lished that changes in gene expression observed upon invitro differentiation reflect developmental regulation duringembryogenesis. The present study compared the expressionof 10 fusion genes generated in mouse ES cells both beforeand after transmission into the germ line. The main findingsof this study are: (a) gene expression in undifferentiated EScells largely reflects the expression in preimplantation em-bryos; and (b) changes in gene expression after ES celldifferentiation in vitro predict highly regulated gene expres-sion during early embryogenesis. Therefore, cultured EScells maintain, to a considerable degree, the transcriptionalspecificity of the pluripotent cells of the preimplantation em-

bryo. Although little is known about the molecular determi-nants of the pluripotent state or the process by which earlystem cells become committed to a specific program of celldifferentiation, these results establish the utility of using cul-tured ES cells to study changes in transcriptional program-ming during cell differentiation and early development.

These results are in general agreement with publishedexpression data for individual genes, assayed by in situ

hybridization. This is perhaps not surprising, given that em-

bryo-derived stem cells are unlikely to maintain totipotency ifthey have undergone extensive changes in their transcrip-tional program. Conversely, loss of stem cell pluripotencycould account for reported differences between in vitro andin vivo patterns of expression (1 5, 27).

The present study establishes the utility of screening mu-tations on the basis of differential expression in cultured EScells. Fusion genes that display regulated expression in vitro

are selected for germline transmission, after which geneexpression is analyzed in transgenic embryos. This processis less laborious than screening chimeras made from uns-elected ES cells, allowing a larger number of mutant clones

to be analyzed. The effort required for germline transmissionis not much greater than that required to examine geneexpression in chimeric embryos. The in vitro screen has otherbenefits: (a) fusion gene expression is easier to study aftergermline transmission than in chimeras, especially early indevelopment. This is important, because genes that are reg-ulated in culture are regulated typically at the onset of gas-trulation; (b) the process reduces the effort expended onclones that are not germline competent; and (c) our datasuggest that repression of genes in totipotent stem cells maybe an important step in early development. Therefore, it willbe important to identify cis- and trans-acting elements thatregulate gene expression during early development. By usingES cells, one is able to study transcriptional control to anextent that would not be possible in mammalian embryos.

Several features of the gene-trapping strategies are wellsuited for studying the transcriptional specificity of the plu-ripotent stem cell: (a) most expressed genes in a cell can betargeted, including weakly expressed genes (32); (b) becauseU3�3geo provirus does not appear to alter regulation of up-stream cellular promoters that drive IacZ expression, a-ga-lactosidase provides a convenient histochemical marker forgene expression (1 6, 1 7, 29); (c) because differentiation perse does not affect IacZ expression (1 7, 29), frgalactosidaselevels are likely to reflect transcriptional control rather thanchanges in RNA stability or processing; and (d) cells recov-ered after gene trap selection contain a reporter gene ex-pressed from the natural promoter of the gene and may beused to study factors that regulate transcription.

Early zygotic cell divisions are accompanied by a wide-spread decrease in DNA methylation (1 9, 20). As a result, thegenomes of blastula-stage embryos and embryo-derivedstem cells are hypomethylated. This raises the possibilitythat all genes may be derepressed because of a globalactivation of the genome. Our results argue against thishypothesis, at least at the level of the whole genome: (a) wehave found no evidence that cryptic promoters (i.e., promot-ers not associated with genes) are capable of activating U3

gene expression. All gene trap targets that we have charac-terized to date represent transcribed cellular genes8 (1 8, 37);and (b) we find no evidence for gratuitous gene expression inES cells. Thus, the two constitutively expressed fusion genes

Cell Growth & Differentiation 1399

that were transmitted into the germline appeared to be ex-

pressed in all cells in 1 0.5-day embryos (29); (c) of the >40

instances in which U3 gene trap vectors have disrupted

known genes, more than half encode widely expressed pro-

teins.8 Similarly, most U3f3geo fusion genes were expressed

constitutively; and (a) Most of the regulated fusion genes

appear to be expressed appropriately. In half of the blue-to-

white clones, stem cell expression was restricted progres-

sively to a subset of cells derived from the CM, whereas in

the white-to-blue clones, expression was activated during

gastrulation. In short, as assayed by gene entrapment, most

genes are not expressed simply because of cell culture con-

ditions or widespread derepression of genes in stem cells.

Because embryoid bodies lack the cellular organization of

the embryo, genes that participate in embryonic patterning

may not be regulated in vitro (38). However, other genes may

be expressed in pluripotent cells and maintained only in the

appropriate cells later in development. In the absence of

normal cell interactions, such genes are expected to be

repressed during in vitro differentiation. We predict that the

commitment of pluripotent cells to one developmental fate

will be determined, at least in part, by the repression of genes

that would allow cells to pursue alternative fates. Genes of

this type are expected to encode molecules required for

signal transduction, cell-cell interactions, and transcriptional

control and to be repressed in all but a few cell types during

early development. For example, signaling molecules such

as activins and fibroblast growth factor are expressed in the

ICM and later restricted to the mesoderm during gastrulation

(22, 24, 39). Similarly, two genes identified by differential

expression in this study, Eck and Rex-1/Zfp-42, encode a

receptor tyrosine kinase and a zinc finger protein,

respectively.

In conclusion, we report that the regulation of gene ex-

pression in cultured embryo-derived stem cells typically re-

flects transcriptional control during early development.

These data provide a basis for in vitro genetic screens that

use gene traps to identify and disrupt developmentally reg-

ulated genes. As methods are developed to control ES cell

differentiation into particular cell types (8-14), similar ap-

proaches may facilitate the genetic analysis of specific de-

velopmental processes. It should also be possible to screen

for genes regulated by growth factors, such as basic fibro-

blast growth factor and transforming growth factor f3, for

which the endogenous targets have not yet been identified

(40-44).

Materials and MethodsPlasmids. pU3j3geoSupF was derived from pGgTKNeoU3LacZen(-)(29) and ppgeo (17). pGgTKNeoU3LacZen(-) LacZ sequences from C/al

Fig. 4. lacZ expression in wild-type and transgenic embryos. Males het-erozygous for fusion genes were bred to C57BLJ6 females, and blasto-cysts and postirnplantation embryos were isolated, fixed in 2% paraforrn-aldehyde 2% glutaraldehyde, and stained overnight in X-Gal solution.Left, blastocysts; right, implanted embryos. Numbers of stained blasto-cysts: A, top two; B, all three; C, none; D, top two; E, bottom three. A,photograph of a 7.5-day pc embryo. All other implanted embryos shownwere isolated at 8.5 days pc: A, J1D4 (REX-i); B, J3A3 (Eck); C, 7.4.2(Nrd); D, 2.4; and E, J2B4. Control embryos that did not stain are locatedon the right in A, C, and 0, on the left in E, and at top in B.

14� Transcriptional Specificity in Mouse Stem Cells

9 5. Darling and J. Rossant, unpublished data.

to Nhel were replaced by the C/al to Xhol j3geo fragment. The intemalTKneo gene was replaced with the 200-bp Sau3AI fragment of the bac-terial amber suppressor gene, supF (44). The resulting vector contains the

f3geo gene in U3 with the 5’-LacZ sequences identical to those in

pGgTKNeoU3LacZen(-). The 5’ end of LacZ was derived originally frompSDKLacz,9 which contains Shine and Delgamo (45-46) and Kozak (47)

consensus sequences.

Cells and Viruses. Cell lines expressing a packaging-defective eco-tropic helper virus (‘P2) were transfected with pU3f3geoSupF and selected

in 500 pg/mI G4i 8. Titering of producer cell lines was carried out asdescribed previously (48). NIH 3T3 cells were cultured in DMEM supple-

mented with 10% calf serum, 10 units of penicillin/mI, and 10 �ig ofstreptomycin/mI. ES cells were cultured on irradiated mouse embryo

fibroblast layers in DMEM supplemented with 15% fetal bovine serum

(heat inactivated at 55#{176}Cfor 30 mm), ioo m� nonessential amino acids(Life Technologies, Inc.), 0.1 mr�i 2-mercaptoethanol, iO units of penicillin

per ml, 10 ;.Lg of streptomycin per ml, and 1000 units ofleukemia inhibitoryfactor (ESGRO, Ufe Technologies, Inc.) per ml. ES-D3 (129; XY, agoutVagouti) cells derived originally by RoIf Kemler were the gift of Janet

Rossant and Rudolf Jaenisch. To infect ES cells, i x i0� cells were

seeded in 6-cm plates and incubated overnight. Cells were incubated at

37#{176}Cwith 1 ml of appropriately diluted and filtered viral supernatant fromproducer line ‘I’2-f3geo2 (titer on NIH 3T3 cells, 2 x i 0” ned’ colony-

forming units per ml per i06 producer cells) in 8 �g/ml Polybrene for 1 hwith occasional rocking. NeoR clones were selected in ES medium con-taming 300 j.�g/ml G4i8 for iO-i4 days, at which point individual colonieswere picked into 24-well plates and expanded.

IacZ Expression during In VItm Differenflatlon. Embryoid bodieswere generated as described (49). Nearly (80%) confluent 10-cm plates ofES cells were trypsinized lightly and grown in suspension in DMEM

supplemented with 10% FCS, 10 units of penicillin per ml, and iO �g ofstreptomycin per ml. After 5 days, the mixture of simple and cysticembryoid bodies was plated onto gelatinized tissue culture plates andincubated for an additional 4 days. Undifferentiated and differentiatedcells were washed with PBS, fixed in 0.5% glutaraldehyde-PBS for iOmm, and stained with X-Gal solution for 6 h at 37#{176}C,as described previ-

ously (29).RNA Analysis. U3 fusion transcripts were detected by Northern blot

analysis. Ten p.g of total RNA were fractionated on i % formaldehyde-agarose gels and transferred onto nitrocellulose membranes (Scleicherand Schuell). Probes were labeled with [�2P]dATP by the random prime

labeling method (50). The Nec probe was prepared from a 1 .4-kb BamHl

TKneo fragment derived from the plasmid pU3HisTKNeo (48).

Cellular RNA from undifferentiated ES cells and differentiated embryoidbodies was analyzed by ribonucelase protection. Probes complementaryto the provirus coding strand were generated by using T3 RNA polymer-ase to transcribe a 689-nt BamHl-Hpal fragment of pGgTKNeo-U3LacZen(-) cloned into pBluescript KS(-) (Stratagene). This probe in-cludes sequences from env, U3, and lacZ and protects a 501 -nt fragmentfrom 5’ fusion transcripts and a 643-nt fragment from 3’ internal tran-

scripts. In addition, a 280-nt NotI-EcoRV fragment of the 132 large ribo-

somal subunit (Si) was transcribed at 1 0% of the specific activity of thelacZ probe to use as an internal control. Thirty �g of cellular RNA werehybridized to excess 32P-Iabeled probes overnight at 55’C. After hybrid-ization to both probes simultaneously, samples were digested with 5j.�g/ml ANase A and 2 pg/mI RNase Ti and processed for electrophoresis

as described previously (48, 52). Protected fragments were separated on6% polyacrylamide-8.3 M urea gels and visualized by autoradiography.

ConstructIon of Germllne Chimeras. Both individual NeoR ES cellclones and pools of three Neo�R clones were injected into preimplantationC57/BL6 blastocysts (49). Chimeric mice were identified by the presenceof agouti coat color, and males were mated to C57/BL6 females to assess

germlmne transmission. The presence of specific transgenes in F1 progeny

was assayed by Southern blot analysis of DNA isolated from tails (�3).Embryonic Expression of IacZ Fusion Genes. Males heterozygous

for proviral insertions were mated to C57/BL6 females. The presence ofvaginal plugs in females was determined the next morning, and noon ofthat day was defined as day 0.5 of embryonic development. Embryos from

various stages of development were isolated from pregnant females, fixedin PBS-2% paraformaldehyde-0.2% glutaraldehyde for 10-30 mm at 4’C,

and rinsed in PBS for i h at 4’C before staining. Staining was performedovernight in PBS-0.02% NP4O-0.0i % SDS-2 m� MgCI2-5 m�

K3Fe(CN)6-5 mM K4Fe(CN)6-1 mg of X-Gal per ml (pH 7.2). The stainingsolution for day-9.5 and older embryos was prepared in 50 m� Tris-

buffered saline (pH 8.0), to reduce background staining due to endoge-nous lysosomal j3-galactosidase activity. After staining, the embryos were

rinsed in PBS and visualized by dark-field microscopy.

AcknowledgmentsWe thank Sita Reddy for allowing us to include data on the 2E7 clone; Phil

Soriano, Brigid Hogan, and Phil Leader for providing the Ilgeo gene,

8.5-day embryo cDNA library, and L32 probe, respectively; members ofthe Nancy Hopkins and H. Eari Ruley laboratories for helpful discussion;and Brigid Hogan, Mark Boothby, and Geoff Hicks for critical readings of

the manuscript.

References1 . Evans, M. J., and Kaufman, M. H. Establishment in culture of pluripo-tential cells from mouse embryos. Nature (Lond.), 292: 1 54-1 56, 1981.

2. Martin, G. A. Isolation of a pluripotent cell line from early mouseembryos cultured in medium conditioned by teratocarcinoma stem cells.Proc. NatI. Acad. Sci. USA, 78: 7634-7638, 1981.

3. Bradley, A., Evans, M., Kaufman, M. H., and Robertson, E. Formation

of germ-line chimaeras from embryo-derived teratocarcinoma cell lines.Nature (Lond.), 309: 255-256, 1984.

4. Wagner, E., Ruther, U., Muller, A., Stewart, C., Gilboa, E., and Keller, G.Expressing foreign genes in stem cells and mice. Cold Spring HarborSymp. Quant. Biol., 50: 103-i i3, 1985.

5. Robertson, E., Bradley, A., Kuehn, M., and Evans, M. Germ-line trans-mission of genes introduced into cultured pluripotential cells by retroviralvector. Nature (Lond.), 323: 445-448, i 986.

6. Beddington, A. S. P., and Robertson, E. J. An assessment of thedevelopmental potential of embryonic stem cells in the midgestationmouse embryo. Development (Camb.), 105: 733-737, 1989.

7. Dcetschman, T. C., Eistetter, H., Katz, M., Schmidt, W., and Kemler, A.The in vitro development of blastocyst-derived embryonic stem cell lines:

formation of visceral yolk sac, blood islands and myocardium. J. Embryol.

Exp. Morphol., 87: 27-45, i985.

8. Burkert, U., von A#{252}den,T., and Wagner, E. F. Early fetal hematopoieticdevelopment from in v/tm differentiated embryonic stem cells. New Biol.,3: 698-708, i99i.

9. Dekel, I., Magal, Y., Pearson-White, S., Emerson, C. P., and Shani, M.

Conditional conversion of ES cells to skeletal muscle by an exogenousMyoDi gene. New Biol., 3: 217-224, 1992.

io. Snodgrass, H. A., Schmitt, A. M., and Bruyns, E. Embryonic stem

cells and in vitro hematopoiesis. J. Cell. Biochem., 49: 225-230, 1992.

1 i . Keller, G., Kennedy, M., Papayannopoulou, T., and Wiles, M. V. He-matopoietic commitment during embryonic stem cell differentiation inculture. Mol. Cell. Biol., 13: 473-486, 1993.

12. Miller-Hance, W. C., LaCorbiere, M., Fuller, S. J., Lyons, G., Schmidt,

C., Robbins, J., and Chien, K. A. In vitro chamber specification during

embryonic stem cell cardiogenesis. J. Biol. Chem., 268: 25244-25252,i993.

i 3. Nakano, T., Kodama, H., and Honjo, T. Generation of lymphohema-topoletic cells from embryonic stem cells in culture. Science (Washington

DC), 265: i098-iiOi, 1994.

14. Rohwedel, J., Maltsev, V., Bober, E., Arnold, H-H., Hescheler, J., and

Wobus, A. M. Muscle cell differentiation of embryonic stem cells reflectsmyogenesis in vivo: developmentally regulated expression of myogenic

determination genes and functional expression of ionic currents. Dev.Biol., 164: 87-ioi, i994.

i5. Poirier, F., Chan, C-T. J., Timmons, P. M., Robertson, E. J., Evans, M.

J., and Rigby, P. W. J. The murine H19 gene is activated during embryonicstem cell differentiation in vitro and at the time of implantation in thedeveloping embryo. Development (Camb.), 1 13: 1 1 05-1 1 14, 1991.

Cell Growth & Differentiation 1401

16. Gossler, A., Joyner, A. L., Rossant, J., and Skames, W. C. Mouse

embryonic stem cells and reporter constructs to detect developmentallyregulated genes. Science (Washington DC), 244: 463-465, 1989.

1 7. Friedrich, G., and Soriano, P. Promoter traps in embryonic stem cells:

a genetic screen to identify and mutate developmental genes in mice.Genes Dev., 5: 15i3-1523, i991.

18. von Melchner, H., DeGregori, J., Rayburn, H., Friedel, C., Reddy, S.,and Auley, H. E. Selective disruption of genes expressed in totipotent

embryonal stem cells. Genes Dev., 6: 919-927, 1992.

19. Monk, M., Boubelik, M., and Lehnert, S. Temporal and regional

changes in DNA methylation in the embryonic, extraernbryonic and germcell lineages during mouse embryo development. Development (Camb.),

99: 371-382, 1987.

20. Kafri, T., Ariel, M., Brandeis, M., Shemer, A., Urven, L, McCarrey, J.,Cedar, H., and Razin, A. Developmental pattern of gene-specific DNAmethylation in the mouse embryo and germ line. Genes Dev., 6: 70�-714,

1992.

21 . Alouani, S., Ketchum, S., Rambosson, C., and Eistetter, H. A. Tran-scriptional activity of the neuron-specific enolase (NSE) promoter in mu-rine embryonic stem (ES) cells and preimplantation embryos. Eur. J. Cell

Biol., 62: 324-332, 1993.

22. Niswander, L., and Martin, G. A. Fgf-4 expression during gastrulation,

myogenesis, limb and tooth development in the mouse. Development

(Camb.), 114: 755-768, 1992.

23. Yamada, G., Kioussi, C., Schubert, F. A., Eto, Y., Chowdhury, K.,

Pituello, F., and Gruss, P. Regulated expression of Brachyury (7), NKX1. 1and PAX genes in embryoid bodies. Biochem. Biophys. Res. Commun.,

199: 552-563, 1994.

24. Ma, Y-G., Rosfjord, E., Huebert, C., Wilder, P., Tiesman, J., Kelly, D.,

and Rizzino, A. Transcriptional regulation of the murine k-FGF gene in

embryonic cell lines. Dev. Biol., 154: 45-54, 1992.

25. Nishi, M., Kumar, N. M., and Gilula, N. B. Developmental regulation of

gap junction gene expression during mouse embryonic development.Dev. Biol., 146: 117-130, 1991.

26. Rogers, M. B., Hosler, B. A., and Gudas, L J. Specific expression of

a retinoic acid-regulated, zinc-finger gene, Rex-i, in preimplantation em-

bryos, trophoblast and spermatocytes. Development (Camb.), 1 13: 8i 5-824, 1991.

27. Poirier, F., Timmons, P. M., Chan, C-T., Guenet, J-L, and Rigby, P. W.

J. Expression of the Li 4 lectin during mouse embryogenesis suggestsmultiple roles during pre- and post-implantation development. Develop-

ment (Camb.), 115: 143-155, 1992.

28. Wurst, W., Rossant, J., Prideaux, V., Kownacka, M., Joyner, A., Hill, D.

P., Guillemot, F., Gasca, S., Cado, D., Auerbach, A., and Ang, S-L. Alarge-scale gene-trap screen for insertional mutations in developmentallyregulated genes in mice. Genetics, 139: 889-899, 1995.

29. Reddy, S., Rayburn, H., von Melchner, H., and Ruley, H. E. Fluores-

cence-activated sorting of totipotent embryonic stem cells expressing

developmentally regulated lacZ fusion genes. Proc. NatI. Acad. Sci. USA,

89: 6721-6725, i992.

30. Reddy, S., DeGregori, J., von Melchner, H., and Auley, H. E. Retro-virus promoter-trap vector to induce lacZ gene fusions in mammaliancells. J. Virol., 65: 1507-1515, 1991.

31 . Mann, A., Mulligan, A. C., and Baltimore, D. Construction of a retro-virus packaging mutant and its use to produce helper-free defectiveretrovirus. Cell, 33: 1 53-1 59, 1983.

32. Chang, W., Hubbard, C., Friedel, C., and Ruley, H. E. Enrichment ofinsertional mutants following retrovirus gene trap selection. Virology, 193:

737-747, 1993.

33. Chen, J., Nachabeh, A., Scherer, C., Ganju, P., Reith, A., Bronson, A.,

and Auley, H. E. Germ-line inactivation of the murine Eck receptor tyrosmnekinase by gene trap retroviral insertion. Oncogene, 12: 979-988, 1996.

34. Hosler, B. A., LaRosa, G. J., Grippe, J. F., and Gudas, L. Expression

of REX-i , a gene containing zinc finger motifs, is rapidly reduced byretinoic acid in F9 teratocarcinoma cells. Mol. Cell. Biol., 9: 5623-5629,1989.

35. Ganju, P., Shigemoto, K., Brennan, J., Entwistle, A., and Reith, A. D.The Eck receptor tyrosine kinase is implicated in the pattern formation

during gastrulation, hindbrain segmentation and limb development. On-

cogene 9: 1613-i624, i994.

36. Ruiz, J. C., and Robertson, E. J. The expression of the receptor-protein tyrosine kinase gene, eck is highly restricted during early mouse

development. Mech. 0ev., 46: 87-100, 1994.

37. DeGregori, J. V., Russ, A., von Melchner, H., Rayburn, H., Priyaranjan,

N., Jenkins, N., Copeland, N., and Ruley, H. E. A murine homolog of the

yeast RNA1 gene is required for post-implantation development. Genes

Dev., 8: 265-276, 1994.

38. Rossant, J., and Hopkins, N. Of fin and fur: genetic analysis ofvertebrate embryonic development. Genes Dcv., 6: 1-i 3, 1992.

39. Albano, A. M., Groome, N., and Smith, J. C. Activins are expressed in

preimplantation mouse embryos and in ES and EC cells and are regulated

on their differentiation. Development (Camb.), 117: 711-723, i993.

40. Forrester, L. M., Nagy, A., Sam, M., Watt, A., Stevenson, L, Bernstein,

A., Joyner, A. L, and Wurst, W. An induction gene trap screen in embry-

onic stem cells: identification of genes that respond to retinoic acid invitro. Proc. NatI. Acad. Sci. USA, 93: 1677-i682, i996.

41 . Kessel, M., and Gruss, P. Murine developmental control genes. Sci-

ence (Washington DC), 249: 374-379, 1990.

42. Jessell, T. M., and Melton, D. A. Diffusible factors in vertebrate de-

velopment. Cell, 68: 257-270, 1992.

43. Kessler, D. S., and Melton, D. A. Vertebrate embryonic induction:mesodermal and neural patterning. Science (Washington DC), 266: 596-

604, 1994.

44. Reik, W., Weiher, H., and Jaenisch, A. Replication-competent Mob-

ney murine leukemia virus carrying a bacterial suppressor tANA gene:

selective cloning of proviral and flanking host sequences. Proc. NatI.Acad. Scm. USA, 82: ii4i-ii4S, 1985.

45. Shine, J., and Dalgarno, L The 3’ terminal sequence of Escherichiacoil 165 ribosomal ANA: complementarity to nonsense triplets and ribo-some binding sites. Proc. NatI. Acad. Scm. USA, 71: i342-1346, 1974.

46. Steitz, J. A., and Jakes, K. A. How ribosomes select initiator regionsin mANA: base pair formation between the 3’ terminus of i6S rANA and

the mANA during initiation of protein synthesis in Escherichia coil. Proc.Natl. Acad. Sci. USA, 72: 4734-4738, i97S.

47. Kozak, M. Point mutations define a sequence flanking the AUG initi-ator codon that modulates translation by eukaryotic ribosomes. Cell, 44:

283-292, 1986.

48. von Melchner, H., and Ruley, H. E. Identification of cellular promoters

by using a retrovirus promoter trap. J. Virol., 63: 3227-3233, 1989.

49. Robertson, A. J. Teratocarcinomas and Embryonic Stem Cells: APractical Approach. Practical Approach Series. New York: Oxford Univer-

sity Press, 1987.

50. Feinberg, A. P., and Vogelstein, B. A technique for radiolabeling DNArestriction endonuclease fragments to high specific activity. Anal. Bio-chem., 137: 266-267, i984.

Si . Shen, M. M., and Leder, P. Leukemia inhibitoryfactor is expressed bythe preimplantation uterus and selectively blocks primitive endoderm

formation in vitro. Proc. NatI. Acad. Sci. USA, 89: 8240-8244, 1992.

52. von Melchner, H., Reddy, S., and Auley, H. E. Isolation of cellularpromoters by using a retrovirus promoter trap. Proc. NatI. Acad. Sci. USA,

87: 3733-3737, 1990.

53. Laird, P. W., Zijderveld, A., Linders, K., Rudnicke, M. A., Jaenisch, A.,

and Berns, A. Simplified mammalian DNA isolation procedure. NucleicAcids Aes., 19: 4293., i99i.