Complex organization and structure of sense and antisense transcripts expressed from the DIO3 gene...

www.elsevier.com/locate/ygeno

Genomics 83 (2004) 413–424

Complex organization and structure of sense and antisense transcripts

expressed from the DIO3 gene imprinted locus$

Arturo Hernandez,* Maria E. Martinez, Walburga Croteau, and Donald L. St. Germain

Department of Medicine, Dartmouth Medical School, Lebanon, NH 03756, USA

Received 25 April 2003; accepted 22 August 2003

Abstract

The human DIO3 gene and its mouse homolog, Dio3, map to chromosomes 14q32 and 12F1, respectively, and code for the type 3

deiodinase, an enzyme that inactivates thyroid hormones and is highly expressed during pregnancy and development. Mouse Dio3 is

imprinted and preferentially expressed from the paternal allele in the fetus. We analyzed the human DIO3 genomic region and identified a

gene (DIO3OS) that is transcribed in the antisense orientation. Multiple DIO3OS transcripts are expressed in most tissues. The structure of

several DIO3OS cDNAs obtained by RT-PCR-based techniques reveals the occurrence of numerous splice variants. The exon–intron

structures of DIO3OS are similar in mouse and human, but the homology of the exonic sequence is very low, except for the first exon, and no

conserved open reading frame is present. We also detected DIO3 transcripts containing additional 5Vuntranslated sequence and a potential

alternative upstream promoter for mouse Dio3. Exonic sequence of a Dio3os cDNA overlaps with the Dio3 promoter and strong promoter

activity in the antisense orientation is detected in a genomic fragment located 3Vof mouse and human DIO3 but not in the DIO3 promoter

region. These results suggest that the DIO3 gene may lie within the structure of the antisense gene, a complex arrangement often observed in

imprinted loci.

D 2003 Published by Elsevier Inc.

Keywords: DIO3 promoter; Type 3 deiodinase; Antisense gene; Genomic imprinting

The Dio3 gene codes for the type 3 deiodinase (D3), a respectively) have been localized to mouse chromosome

selenoprotein that plays a central role in thyroid hormone

metabolism [1]. D3 enzymatic activity results in the inacti-

vation of thyroid hormones, since it transforms both the

prohormone T4 and the active hormone T3 into metabolites

that are biologically inactive [1]. D3 displays a marked

developmental pattern of expression: it is highly expressed

in pregnant uterus [2], placenta [3,4], and the fetus [5], while

in the adult, expression is low and limited to a few selected

tissues [6,7]. D3 also is strongly induced by serum, growth

factors, and phorbol esters in cell culture systems [8,9].

D3 cDNAs have been isolated from several species [10–

13] and the mouse and human D3 genes (Dio3 and DIO3,

0888-7543/$ - see front matter D 2003 Published by Elsevier Inc.

doi:10.1016/j.ygeno.2003.08.024

$ Sequence data from this article have been deposited with the EMBL/

GenBank Data Libraries under Accession Nos. AF469199–AF469208,

AY077457–AY077459, AY283181, AY283182, and W97869.

* Corresponding author. Department of Medicine, Dartmouth Medical

School, Borwell Building, Room 720 West, One Medical Center Drive,

Lebanon, NH 03756. Fax: +1-603-650-6130.

E-mail address: [email protected] (A. Hernandez).

12F1 and human 14q32 [14]. Characterization of the mouse

Dio3 gene has shown that the coding and 3V-untranslatedregions are contained in a single exon, approximately 1.9

kb in length (GenBank Accession No. AF426023), which is

transcribed using a promoter located immediately upstream

[15]. The expression of the Dio3 gene results in a transcript

of 2.1 kb in size that is most abundant in those rat and

mouse tissues showing the highest D3 activity, such as

decidual tissue [2] and placenta [11], and in growth factor-

stimulated cell culture systems [8,9]. However, larger tran-

scripts have also been detected in the brain of hyperthyroid

adult rats, when a rat D3 cDNA is used as a probe [16].

Interestingly, the mouse Dio3 gene is imprinted and

preferentially expressed from the paternal allele in the

developing fetus [17,18]. To characterize the DIO3 and

Dio3 loci further, we have identified and partially char-

acterized both human and mouse genes that overlap with

the DIO3 and Dio3 promoter regions and are transcribed

in an antisense orientation. We have designated these

A. Hernandez et al. / Genomics 83 (2004) 413–424414

human and mouse genes DIO3OS and Dio3os, respec-

tively, and speculate that they may play a role in the

maintenance of Dio3 preferential monoallelic expression.

Results

Identification and structure of human DIO3OS

About 10 kb of the mouse and human DIO3 5Vflanking regions was subcloned from previously isolated

P1 genomic clones [14] and sequenced, and their sequen-

ces were compared. The first 1.5 kb of 5V flanking region

was found to be extremely G+C rich (80% of the

sequence) and was highly conserved between mouse

and human (83% homology, Fig. 1). Further 5V thehomology is low (less than 60%), except for a 180-bp

region that is 80% homologous, contains a conserved

BamHI site, and is located approximately at positions

Fig. 1. Structure of the human DIO3OS gene. The exon– intron structures,

compared to the genomic sequence, of 13 cDNAs for the DIO3OS gene are

shown. The conserved short region containing the proximal polyadenyla-

tion site is indicated in black and circled in the genomic diagram. Exons are

depicted as thick bars and introns as thin bars. The first 3 cDNAs were

found in GenBank (accession numbers shown), while the other 10

(numbered from 1 to 10) were generated in these studies from RNA of

different human tissues by RT-PCR and RACE methods. The specific

method and human RNA samples used to generate cDNAs 1–10, along

with their GenBank accession numbers, are listed for each cDNA under

Materials and methods. The six exons most consistently observed in the

DIO3OS gene are designated at the bottom as 1 through 6.

�4200 and �4000 from the human and mouse D3

transcription start sites, respectively (circled in Fig. 1).

Mouse and human genomic sequences containing this

short, conserved region were found to match the sequen-

ces of several mouse and human expressed sequence tags

deposited in GenBank (Fig. 1). All these human and

mouse ESTs contained a consensus polyadenylation site

(referred to as the proximal site) and a poly(A)+ tail at

their 3V end. The orientation of their 3V ends indicated that

they were transcribed in an orientation antisense to that of

DIO3. The 180 bp of conserved sequence was found to

contain and flank the consensus polyadenylation site.

Further analysis identified another group of human ESTs

in GenBank that matched a human genomic region

located 9.25 kb upstream of the DIO3 start site (Fig.

1). This second group of ESTs also contained putative

polyadenylation sites (referred to as the distal site) and

their 3V ends were oriented in the same direction, that is,

antisense with respect to DIO3. In the mouse, no addi-

tional ESTs were found to match a similar, distal genomic

region. These data showed that at least one other gene,

which we have designated in the human and the mouse

DIO3OS and Dio3os, respectively (OS for opposite

strand), is transcribed from this locus in an orientation

that is antisense to that of DIO3 and Dio3.

Three human ESTs for DIO3OS deposited with GenBank

(GenBank Accession Nos. AI290455, R73460, and

AF305836) that contain the proximal polyadenylation site

were sequenced. Comparison of these data with the genomic

sequence revealed the occurrence of alternative splicing

(Fig. 1, top three cDNAs). The most 5V sequence of these

cDNAs corresponds to an exon (designated as 1) that shows

96% conservation between mouse and human and is located

1 kb upstream of the DIO3 transcription start, within the

G+C-enriched region (Fig. 1).

To characterize the exon–intron structure of DIO3OS

further and to clarify whether both human proximal and

distal polyadenylation sites belong to the same gene, we

performed 3V and 5V rapid amplification of cDNA ends

(RACE) and RT-PCR using RNA from several human

tissues. Several DIO3OS cDNAs were obtained, subcl-

oned, sequenced, and assigned a number for use in this

article. The structures of these cDNAs are shown in Fig. 1,

in which, at the bottom, we have numbered the exons most

consistently observed from 1 to 6. The structures of these

cDNAs reveal a complex splicing pattern. The sequence of

the five cDNAs at the top of Fig. 1 shows alternative

splicing of introns between exons 2, 3, and 4 and within

exon 4. The structures of DIO3OS cDNAs 6 to 10

demonstrate that the distal polyadenylation site is also

used for transcription from the DIO3OS gene. A very short

variant of exon 4 is seen in cDNA 6, while up to three

different splicing variants of exon 5 are observed in

cDNAs 5, 8, and 9. This variable pattern of splicing leads

to a complex mix of transcripts encoded by the DIO3OS

gene.

enomics 83 (2004) 413–424 415

Expression pattern of human DIO3OS

We used human DIO3OS cDNA 1 as a probe to analyze

several human multiple-tissue Northern blots. The results

are shown in Fig. 2 and reveal that the human DIO3OS gene

is widely expressed in many tissues. A complex pattern of

transcripts is observed. The most prominent bands are

approximately 4.8, 3.7, 2.6, and 1.4 kb in size and are

detected in most tissues. Other bands, some of them very

small and some of them appearing to be doublets, are also

observed, e.g., in pancreas (Figs. 2A, lane 8, and 2C, lane

1), heart (2B, lane 3), and muscle (2B, lane 8). A weaker

band corresponding to a very large transcript (approx 8 kb)

is observed in fetal lung (2D, lane 3). DIO3OS transcripts

appear to be most abundant in testis and adrenal cortex (2A,

lanes 4 and 5); prostate, bladder, and uterus (2B, lanes 1, 4,

and 7); placenta (2C, lane 6); and fetal lung (2D, lane 3).

Due to the splicing complexity, it is very difficult to

establish which of the isolated cDNAs correspond to a

specific band in the Northern blot, but an estimation of size

suggests that most of these cDNAs are missing between 500

and 1500 bp of 5V sequence. In addition, probing of the

human fetal Northern blot with cDNA 4 (3 kb in length, see

Fig. 1), which comprises most of the ‘‘distal’’ genomic

sequence, resulted in the identification of a single band in

fetal lung of approximately 8 kb (data not shown) that

coincides in size with that observed in Fig. 2D (lane 3).

Taken together, these results suggest that the distal poly-

adenylation site is used to generate larger and less abundant

transcripts, not easily detectable by Northern analysis.

Exon–intron structure of the mouse Dio3os gene

Exons in mouse Dio3os are structurally similar to human

exons 1 to 4. Sequencing of mouse Dio3os cDNAs also

reveals multiple splicing variants and their structures are

summarized in Fig. 3. Interestingly, the first exon of mouse

A. Hernandez et al. / G

Fig. 2. Tissue expression pattern of the human DIO3OS. Commercially available

(GenBank Accession No. AI290455). Blots were autoradiographed for 24 h and th

of mRNA loaded per lane. (A) Lanes 1, stomach; 2, small intestine; 3, thymus; 4, te

1, prostate; 2, stomach; 3, heart; 4, bladder; 5, small intestine; 6, colon; 7, uterus;

placenta; 7, brain; 8, heart. (D) Lanes 1, fetal kidney; 2, fetal liver; 3, fetal lung; 4,

the left side. Ladder for D is indicated at the right.

Dio3os cDNA 6 (see Fig. 3) includes a portion of the mouse

Dio3 gene promoter and exon, oriented in the opposite

direction.

Despite the similarities in exon–intron structure between

mouse and human, the exonic sequence is not highly

conserved (less than 60% homology), except for exon 1

(96%), which is located in the highly conserved G+C

region. Considering the interspecies similarities in the

antisense gene structure between human and mouse, it is

reasonable to expect that an open reading frame would also

show some conservation and be shared—if only partially—

by different cDNAs. No open reading frame meeting these

criteria was found. However, a 408-bp open reading frame

encompassing all or parts of exons 2, 3, and 4 is found in

one of the human cDNAs deposited with GenBank (Acces-

sion No. AF305836).

Multiple transcripts for human DIO3

A 2.1-kb mRNA coding for the D3 protein has been

noted in human placenta [10], in different mouse and rat

tissues, and in cell culture models [2,21]. A transcript of this

size is expected given the location of the known mouse

promoter [15] and the length of the cDNA. However, larger,

less abundant transcripts have been detected by Northern

blot analysis in hyperthyroid rat brain when a rat Dio3 probe

is used [16]. Interestingly, the largest human cDNA for

DIO3 isolated by Salvatore et al. [11] contains 180 bp that

lay upstream of the expected transcription start site (as

mapped previously for the mouse gene) and includes the

TATA box and other promoter elements. We have obtained a

similar cDNA fragment using 5VRACE and human fetal lung

RNA (data not shown). These results in rat and human

suggest the existence of larger transcripts for DIO3.

To address further this issue of multiple DIO3 tran-

scripts, we hybridized the same multiple-tissue human

Northern blots with the larger human D3 cDNA mentioned

multiple-tissue human Northern blots were probed with an antisense cDNA

en hybridized with a cyclophilin cDNA to provide a control for the amount

stis; 5, adrenal cortex; 6, thyroid; 7, adrenal medulla; 8, pancreas. (B) Lanes

8, muscle. (C) Lanes 1, pancreas; 2, kidney; 3, muscle; 4, liver; 5, lung; 6,

fetal brain. A, B, and C are aligned with the same RNA ladder indicated at

Fig. 3. Structure of the mouse Dio3os gene. The alternative splicing of six cDNAs for the mouse Dio3os gene is shown. Exonic sequences are depicted as thick

bars and splicing donor and acceptor sites are connected by lines. The cDNA number is displayed at the same level as the vertices of the corresponding splicing

lines. These cDNAs were isolated in our laboratory, though the splicing pattern of some of them is compatible with that of some of the ESTs deposited in

GenBank databases. Accession numbers are listed under Materials and methods.


above [11], which includes promoter elements. The results

(Fig. 4) show a strong 2.1-kb band in placenta (Fig. 4C,

lane 6), fetal liver (4D, lane 2), and uterus (4B, lane 7). In

addition, two other bands, approximately 4.8 and 3.2 kb,

were found in several other tissues. The 4.8-kb band was

the most abundant in heart (4B, lane 3, and 4C, lane 8) and

skeletal muscle (lane 8 in 4A and 4B), while the 3.2-kb

band was predominant in testis (4A, lane 4), bladder, and

uterus (4B, lanes 4 and 7). All or some of these three bands

were present in other tissues as well, such as adrenal cortex

(4A, lane 5), thyroid (4A, lane 6), prostate and stomach

(4B, lanes 1 and 2), pancreas (4C, lane 1), and fetal lung

(4D, lane 3).

We also probed a different human Northern blot with a

DIO3 cDNA fragment comprising only the last 200 bp of

the coding region plus all the 3Vuntranslated region. In this

case we detected again bands of 3.2 and 2.1 kb in size, plus

Fig. 4. Expression pattern of human DIO3 transcripts. (A–D) Multiple-tissue hum

Salvatore et al. [11], which contains 180 bp of the 5Vproximal promoter region

autoradiographed for 3 days. Blots and lanes are the same as in Fig. 2. (E) A multip

that resulted from PstI/ApaI digestion of the DIO3 cDNA and includes the

autoradiographed for 3 days. It was then hybridized with a cyclophilin cDNA to pr

heart; 3, kidney; 4, spleen; 5, liver; 6, colon.

an additional 2.5-kb band in heart (Fig. 4E). However, no

4.8-kb transcript was detected, suggesting that this larger

band is not a true D3-coding transcript, since it hybridizes

only with the most 5V region of the larger D3 cDNA.

Interestingly, the other human transcripts detected (2.5 and

3.2 kb in size) match the size of the larger Dio3 transcripts

previously detected in hyperthyroid rat brain [16], indicating

that they could be conserved D3-coding transcripts that

contain an additional 5Vuntranslated region and are possibly

transcribed from an alternative promoter.

Mouse and human D3 promoters are conserved

As noted, the G+C-rich region upstream of the DIO3

gene spans about 1.5 kb and is 83% homologous between

mouse and human. A finer comparative analysis of the

sequence shows that there are three smaller regions, 240,

an Northern blot analysis using as a probe the human D3 cDNA isolated by

. The same blots from Fig. 2 were stripped and hybridized. Blots were

le-tissue human Northern blot was probed with a partial 1.1 kb DIO3 cDNA

end of the coding region and almost all of the 3V-UTR. The blot was

ovide a control of the amount of mRNA loaded per lane. Lanes: 1, brain; 2,

A. Hernandez et al. / Genomics 83 (2004) 413–424 417

420, and 290 bp in size and designated A, B, and C (Fig. 5),

that are conserved to an even higher degree (86, 92, and

96% homology, respectively). Regions A and C (Fig. 5A)

coincide, respectively, with the previously characterized

mouse Dio3 promoter [15] and exon 1 of the Dio3os gene.

Comparison of the mouse and human basal promoter

regions (240 bp of 5V flanking region) shows conservation

of several promoter elements (Fig. 5), including the TATA

box, which is critical for activity since we have previously

demonstrated that its deletion results in loss of 98% of the

Dio3 promoter activity [15].

Fig. 5. (A) Homology between mouse and human D3 genes and promoter regions.

shown in black. Highly conserved sequences within approximately 1.7 kb of 5Vflanrestriction sites are shown. (B) Human DIO3 promoter activity. The promoter ac

region and to the described mouse Dio3 promoter [15] was tested in different cell

galactosidase was cotransfected to correct for transfection efficiency. Bars represe

more different experiments.

Two different human genomic fragments containing the

conserved promoter region show potent promoter activity

(500- to 2000-fold increase over that of the empty vector)

when subcloned in front of a luciferase reporter gene and

transfected into three different cell lines (Fig. 5B). These

activities are similar to those obtained from a mouse

promoter construct under the same conditions (Fig. 5B),

though some differences between cell lines are observed.

Stronger promotion of transcription is observed in cell lines

expressing endogenous Dio3 (XTC-2 and BVS-1 cells) and

certain promoter specificity is observed between species.

The DIO3 exon is shown as a thick bar and the open reading frame (ORF) is

king region (designated A, B, and C) are shown in thinner, gray bars. Not all

tivity of genomic fragments corresponding to the human DIO3 5Vflankinglines by transient transfection experiments. A control plasmid expressing h-nt the means F standard deviations of three different cultures from two or


For example, in rat BVS-1 cells, a fragment of the human

DIO3 promoter has one-third of the promoter activity shown

by a mouse Dio3 promoter fragment of similar length.

DIO3OS 5V-end and promoter analysis

Despite several attempts to isolate and clone the 5Vend of

the DIO3OS gene, we have been unable to extend the 5Vsequence above exon 1 by more than 70 bp. Apparently, the

extremely high G+C content of this genomic region makes

this very difficult when using various RT-PCR-based meth-

ods, probably due to strong secondary structures in both the

RNA and the first-strand cDNA.

Fig. 6. Promoter analysis of the human DIO3 locus. (A) Location and orientation o

by arrows. (B) Promoter activity of the genomic fragments indicated in A. Activiti

under Materials and methods. Results are expressed as a percentage of the activity

control (absolute activities of the DIO3 promoter are shown in Fig. 5B). The data

triplicate cultures from one to three different transfection experiments. The numbe

determined.

To investigate further the possibility that larger Dio3

mRNAs are transcribed by an alternative upstream promoter

and to locate a genomic region that may contain the promoter

of the Dio3os gene, we used available software (bimas.dcrt.

nih.gov/molbio/proscan/ and www.molbiol.ox.ac.uk/) to an-

alyze 20 kb of mouse and human genomic sequence (cen-

tered at the Dio3 gene). Three regions were predicted as

potential sense or antisense promoters: the CG-rich region 5Vof Dio3 and two other regions located within 6 kb 3Vof theDio3 gene. We subcloned several human genomic fragments

that corresponded to these regions and analyzed them for

promoter activity using a luciferase reporter vector trans-

fected into various cells lines (Fig. 6). Several human frag-

f the genomic fragments that were tested for promoter activity are indicated

es were assayed by transient transfections in the cells indicated as described

measured for the DIO3 promoter (fragment 1), which was used as a positive

represent the means F standard deviations of determinations performed in

rs correspond to the genomic fragments indicated by arrows in A. N.D., not

http:\\bimas.dcrt.nih.gov\molbio\proscan\

http:\\www.molbiol.ox.ac.uk\


ments showed significant promoter activity (Fig. 6B). A

genomic region approximately 6 kb 3Vof the DIO3 gene and

centered around conserved KpnI and XbaI restriction sites

(fragments 2 and 8) had strong sense and antisense promoter

activities. These were very significant in more than one cell

line, if we consider that the fragment used as positive control

(the DIO3 promoter, fragment 1) induces a hundreds or

thousands fold increase in promoter activity as shown in

Fig. 5. Shorter fragments from this region lose their promoter

activity (fragments 5, 6, and 7). Also, significant sense

promoter activity is found in a short fragment (No. 3) located

Fig. 7. Promoter analysis of the mouse Dio3 locus. (A) Location and orientation o

by arrows. (B) Promoter activity of the genomic fragments indicated in A. Activiti

under Materials and methods. Results are expressed as a percentage of the activity

control (absolute activities of the Dio3 promoter are shown in Fig. 5B). The data

triplicate cultures from one to three different transfection experiments. The numbe

determined.

5Vof the DIO3 exon. No promoter activity was detected in

COS-7 cells when a fragment containing the DIO3 promoter

in the antisense orientation (No. 10) was transfected.

Strong promoter activity was also detected in various

mouse genomic fragments (Fig. 7). Interestingly, a fragment

corresponding to the region located 3V of Dio3 (No. 9)

displayed antisense promoter activity that was higher than

that measured for the Dio3 promoter (No. 1) when trans-

fected into COS-7 cells. The promoter activity of this

fragment was also significant in BeWo cells. As with the

human locus, a fragment (No. 2) from this region also

f the genomic fragments that were tested for promoter activity are indicated

es were assayed by transient transfections in the cells indicated as described

measured for the Dio3 promoter (fragment 1), which was used as a positive

represent the means F standard deviations of determinations performed in

rs correspond to the genomic fragments indicated by arrows in A. N.D., not


displayed strong promoter activity in the sense orientation,

whereas shorter antisense oriented fragments (Nos. 8 and 10)

did not promote transcription. A genomic fragment located

upstream of the Dio3 exon (No. 4) showed strong promoter

activity in the sense orientation in COS-7 and BeWo cells,

but none in the D3-expressing BVS-1 and XTC-2 cells. No

antisense promoter activity was detected in a fragment (No.

7) corresponding to the Dio3 5V flanking region. Taken

together these results support the data obtained by Northern

analysis in this and previous work [16] and suggest that there

could be an upstream alternative promoter for Dio3. These

results also suggest the possibility of an antisense promoter

in the region located 3Vof Dio3 that may be the origin of the

Dio3os transcripts. Analysis of this region revealed two

groups of sequences, 170 and 220 bp in length, that were

homologous (84 and 81%) between mouse and human

(regions P1 and P2 in Fig. 8A). Region P1 contains putative

AP-1 and serum-response elements, while region P2 con-

tains a putative TATA box. Preliminary results indicate that

Fig. 8. (A) Conserved sequences in the region with antisense promoter activity.

regions are shown in gray and conserved sequences are boxed. Arrows point to a c

(B) Northern blots of RNA from BVS-1 cells probed with the three different genom

hybridizing with probes 1 and 3 that, based on this analysis, include sequences t

the serum-response element in P1 is functional. P1 and P2

regions are not able to promote transcription independently

(see promoter activity of mouse fragments 3, 8, and 10 in

Fig. 7 and human fragments 5, 6, and 7 in Fig. 6), but together

they display robust promoter activity when oriented in either

direction (mouse fragments 2 and 9 in Fig. 7 and human

fragments 2 and 8 in Fig. 6).

No ESTs have been found to match genomic sequences

between the DIO3 gene and the presumed antisense pro-

moter, nor has it been possible to isolate a partial cDNA for

the antisense gene by connecting exon 1 with sequences for

which the strong antisense promoter activity was found.

This is probably due to the difficulties in performing reverse

transcription and PCR through the highly C+G-enriched

DIO3 5Vflanking region. However, we performed Northern

analysis of RNA isolated from Dio3-expressing BVS-1 cells

with three mouse genomic fragments. The first was a

BamHI/KpnI 3.5-kb restriction fragment (Fig. 8B, top left)

that comprised most of the known exon–intron structure of

The region displaying antisense promoter activity is expanded; conserved

onserved putative serum-response element (SRE) and a putative TATA box.

ic fragments indicated in the map. Arrows point to two antisense transcripts

hat are located at both sides of the Dio3 gene.


the antisense gene, the second fragment contained the Dio3

coding region, and the third was a 0.75-kb BglII restriction

fragment located 3Vof the Dio3 gene (Fig. 8B, top). The first

probe showed three major transcripts for the antisense gene

in these cells, approximately 3.5, 1.6, and 1.2 kb in size.

(Note that the relative abundance of these transcripts may

not be accurately judged from this blot since the genomic

probe used likely favors hybridization with larger, less-

spliced transcripts.) As expected, the 2.2-kb transcript for

the Dio3 mRNA is observed when using the second probe.

Interestingly, the third probe revealed two weak transcripts

that matched the size of those obtained with the first probe,

suggesting that short exonic sequences for at least some of

the antisense transcripts are located downstream of the Dio3

gene. This supports further the hypothesis that antisense

exonic and promoter sequences are located 3Vof Dio3.

Discussion

The DIO3 gene, located on human chromosome 14q32,

codes for the type 3 deiodinase, an enzyme highly expressed

in the pregnant uterus, the placenta, and the fetus. We have

previously characterized the mouse Dio3 gene, demonstrat-

ed that all the coding and 3V untranslated regions are

contained in a single exon, and identified its promoter

[15]. Here we characterize the human DIO3 gene and show

that the structures of the gene and promoter elements are

essentially identical to those of the mouse. In addition we

describe other mouse and human ESTs that correspond to

genomic sequences within the DIO3 locus. The location of

the poly(A)+ tail of these ESTs indicates that they are

transcribed antisense of the DIO3 gene and belong to a

second gene at this locus that has been designated DIO3OS.

The DIO3OS gene is expressed in most tissues and

multiple transcripts are detected by Northern analysis. The

structures of these ESTs and various cDNAs generated by

RT-PCR and RACE techniques show the presence of at least

six exons and two alternate polyadenylation sites for human

DIO3OS. Exons 1 to 4 and the proximal polyadenylation

site are also found in mouse Dio3os. The distal polyadeny-

lation site in human DIO3OS is not found in the mouse, and

Northern analysis using RNA from human tissues indicates

that only very large and/or less abundant transcripts are

derived using this site. The possibility that this human-

specific distal polyadenylation site belongs to a different

gene is unlikely but cannot be entirely excluded.

The sequences of the DIO3OS cDNAs demonstrate a

complex pattern of alternative splicing of transcripts from

this gene; up to 13 different DIO3OS cDNAs have been

identified. Although the exon–intron structure is conserved

between mouse and human, the exonic sequence is not

conserved, except for exon 1 (96% homology), which lies

in the highly conserved G+C-rich region flanking the DIO3

gene. A comparison of the sizes of the bands in the Northern

and DIO3OS cDNAs indicates that additional 5Vsequence has

yet to be identified. However, we have not been able to

extend this 5V sequence, presumably due to the high G+C

content of this region. An open reading frame (408 bp of

length) is found in one of the human DIO3OS cDNAs

(Accession No. AF305836). However, this open reading

frame is not found in the corresponding mouse cDNA and

includes a start codon that shows little homology to the

consensus Kozak sequence [22]. Furthermore, the predicted

polypeptide sequence shows no significant homology with

any entries in the protein databases. These observations

suggest that DIO3OS is a noncoding gene, a finding pro-

posed for several antisense genes present in other imprinted

loci [23]. Alternatively, a coding region for theDIO3OS gene

might lie within the highly conserved exon 1 and additional 5Vsequence yet to be identified.

Although the pattern of transcripts corresponding to the

antisense human gene is more complex than that of DIO3, it

is worth noting that there are overall similarities in the

patterns of tissue expression of both genes. The tissues in

which the antisense transcripts are most abundant corre-

spond largely to those expressing higher levels of DIO3

transcripts such as uterus, testis, and bladder. In this regard,

we have found no significant promoter activity in the

antisense orientation for the DIO3 promoter, and further

analysis of the region suggests that the promoter for the

antisense gene is likely located 3Vof DIO3. This suggests

that both genes are not transcribed from the same promoter.

An alternative explanation for the similar tissue expression

pattern is the existence of common regulatory mechanisms

within the imprinted chromosomal domain.

Because of the complexity of this locus, we have

investigated the genomic structure and the pattern of tran-

scripts from the DIO3 gene. We have shown that a proximal

human DIO3 promoter immediately upstream of the coding

region is conserved between human and mouse with regard

to sequence and location. In addition, larger DIO3 tran-

scripts, 4.8, 3.2, and 2.5 kb in size, can be demonstrated

when using a DIO3 cDNA that contains the coding region

and the proximal DIO3 promoter sequence. The largest

transcript is not detected when using a probe corresponding

to the 3Vhalf of the D3 cDNA. This observation indicates

that the 4.8-kb band is not a D3-coding mRNA. Search of

the human and mouse genomic sequences downstream of

the DIO3 gene and 3VRACE experiments have not detected

any alternative splicing or additional 3V untranslated se-

quence that might correspond to the DIO3 gene (data not

shown). Taken together, these results suggest that the 3.2-

and 2.5-kb transcripts, previously detected in the rat brain

[16], are likely D3-coding mRNAs containing additional 5Vuntranslated sequence and transcribed from an upstream

alternative promoter. In fact, herein we have shown strong

promoter activity of a mouse genomic fragment located 5Vofthe previously characterized proximal promoter [15]. Pro-

moter activity is also significant in a similarly located

human fragment. These fragments contain sequences

corresponding to exons 1 and 2 of the antisense gene in


both species. Along with the isolation of a mouse Dio3os

cDNA that includes sequences from the Dio3 promoter

(cDNA 6), these results indicate that both Dio3 and Dio3os

genes overlap at least in their 5Vregions and that the overlap

occurs in the G+C-rich region. The occurrence of antisense

promoter activity in a region located 3V of Dio3, the

conservation of DNA sequences and putative promoter

features, and the detection of mRNA transcripts of the same

size using genomic probes at both sides of the Dio3 gene

also suggest that the Dio3 gene lies within the structure of

the Dio3os gene.

A genomic arrangement in which the structures of two or

more genes overlap in antisense orientation is rare, but has

been described with imprinted genes [24]. Since the mouse

Dio3 shows preferential expression from the paternal allele

[17,18], this locus might be another of several examples of

imprinted genes that are associated with overlapping anti-

sense transcripts [25–27]. Such antisense transcripts are

usually also imprinted and noncoding [24], have been

proposed to serve regulatory functions [28], and in certain

cases are a critical part of the mechanisms of genomic

imprinting [29]. We speculate that this may also be the case

for the Dio3os gene and its human homolog. More research

is needed to characterize this interesting locus fully and to

determine the role of DIO3OS.

Materials and methods

Genomic clones, restriction mapping, vectors, subcloning

and sequencing

Genomic human and mouse (129SVJ strain) P1 clones for

the DIO3 gene were identified using a PCR-based screening

system as described [14] and purchased from Genome

Systems (St. Louis, MO, USA). cDNA clones corresponding

to expressed sequence tags deposited with GenBank (Ac-

cession Nos. AF305836, AI29045, R73460, AW556460,

and W97869) were obtained from the same company.

Additional cDNAs generated in this work were sequenced

and this information was deposited with GenBank. Their

accession numbers are listed later in this section (see below)

along with the methods used to generate them.

Restriction maps of both human and mouse genomic

clones were obtained by digestion with various restriction

enzymes (Life Technologies, Gaithersburg, MD, USA).

Southern blotting followed standard procedures using the

rat and human D3 cDNAs as probes. Genomic DNA frag-

ments were subcloned into pBluescript SK (Stratagene, La

Jolla, CA, USA) and pXP2 [19]. The pCH110 h-galactosi-dase expression vector was obtained from Pharmacia–Bio-

tech (Piscataway, NJ, USA).

Genomic DNA fragments used in transient transfection

studies were subcloned following standard procedures into

the pXP2 vector [19], which contains a luciferase reporter

gene. PCR products were subcloned using the PCR Script

kit from Stratagene. Sequencing was performed using an

automated sequencer with fluorescent dye terminators (PE

Applied Biosystems, Foster City, CA, USA).

Cell cultures and transient transfections

Human choriocarcinoma BeWo and COS-7 cells were

obtained from the American Type Culture Collection (Man-

assas, VA, USA). BVS-1 cells are a D3-expressing cell line

we have generated by subculturing precursor cells from rat

brown adipose tissue [20]. Both BeWo and BVS-1 cells were

cultured in high-glucose DMEM supplemented with 10%

fetal bovine serum (Life Technologies). COS-7 cells were

grown in the same medium supplemented with 5% fetal

bovine serum and 5% horse serum. Xenopus laevis XTC-2

and XL-2 cell lines were kindly provided by Dr. J. Tata

(National Institute for Medical Research, The Ridgeway,

London, UK) and cultured in 0.6� Leivobitz-15 medium

(Sigma, St. Louis, MO, USA) supplemented with 10% fetal

bovine serum (Life Technologies). All cell types were plated

at a density of approximately 2 � 105 cells per well in six-

well culture dishes. Cells were transfected the next day using

the FuGene6 reagent (Roche, Indianapolis, IN, USA) with

1–2 Ag of plasmid DNA and 0.5 Ag of h-galactosidasepCH110 expression vector (Pharmacia–Biotech) to correct

for transfection efficiency. Luciferase and h-galactosidaseactivities were determined in cell lysates using assay

reagents from Promega Corp. (Madison, WI, USA). Light

emission was quantified using an EG&G Berthold micro-

plate luminometer LB 96V (Wallac, Gaithersburg, MD).

Northern analysis

Multiple-tissue human Northern blots were obtained from

Clontech (Palo Alto, CA, USA) and Origene Technologies

(Rockville, MD, USA). Blots were prehybridized at 42jCfor 8–20 h in 50% formamide, 5� SSC, 5� Denhardt’s, 50

mM sodium phosphate, pH 6.5, and 0.2% SDS and then

hybridized at 42jC overnight with the specified radioactive

probe in the same buffer containing 2� Denhardt’s. Blots

were washed four times for 15 min at room temperature with

2� SSC/0.1% SDS and twice for 20 min with 0.1� SSC/

0.1% SDS at 50 to 65jC and then autoradiographed. cDNAs

and genomic DNA fragments were labeled with radioactive

[32P] dCTP (ICN Biochemicals, Inc., Costa Mesa, CA, USA)

using the Oligolabelling Kit (Pharmacia), and probes were

purified through G-50 columns (Pharmacia). Total RNA

from BVS-1 cells was isolated using a guanidinium chlo-

ride-based method as previously described [8].

Generation of cDNAs by RT-PCR and RACE

DIO3OS and Dio3os cDNAs were generated by 5VRACE,by PCR using different human Marathon Ready cDNA kits

(Clontech), or by RT-PCR of mRNAs obtained from Clon-

tech or in our laboratory. ExTaq polymerase (Panvera Corp.,


Madison, WI, USA) was used in all PCR reactions. Reverse

transcriptase (Superscript II) was obtained from Life Tech-

nologies. Primer sequences are listed below. The technique

and tissue sample used, GenBank accession numbers, and

primers utilized to generate the specific cDNAs described in

this work and numbered in Figs. 1 and 3 are as follows:

human DIO3OS cDNAs 1 (GenBank AF469199) and 2

(GenBank AF469200) were obtained from human placenta

by PCR using primers hc5low1 and hc5100; human

DIO3OS cDNAs 3 (GenBank AF469201), 4 (GenBank

AF469202), and 5 (GenBank AF469203) were obtained

by 5VRACE from human fetal lung using the primer NGas2;

human DIO3OS cDNA 6 (GenBank AF469204) was

obtained from fetal lung by PCR using primers NG10 and

hc51500; human DIO3OS cDNA 7 (GenBank AF460205)

was obtained from fetal lung by PCR using primers NG8

and hc5120s; human DIO3OS cDNAs 8 (GenBank

AF469206), 9 (GenBank AF469207), and 10 (GenBank

AF469208) were obtained by RT-PCR from human liver

using primers hc5low1 and Ngas2. Accession numbers for

mouse Dio3os cDNAs indicated in Fig. 3 are No. 1,

AY077457; No. 2, AY283182; No. 3, AY283181; No. 4,

W97869; No. 5, AY077458; and No. 6, AY077459.

Primers

Primers were obtained from Life Technologies and

their sequences are as follows: NGas2 (antisense), 5V-CTCATGCAACGGAGGCTGAG-3V; hc5-100 (antisense),

5 V-ATCTGGGCCAGCCTGGGAG-3 V; hNG8, 5 V-CAAGCGGAGGACACTGCGG-3V; NG10 (antisense),

5V-CTGCGGCTGTACTTTGTCCCC-3V; hc5low1 (sense),

5V-CTGAGCCACCTTCGCGGAC-3V; hc5120s (sense), 5V-GTTTGCCCAGCAGACCTCCC-3V; hc51500 (sense), 5V-GCCCAATAGGAAGCACCTG-3V; mST75 (sense), 5V-CA-GAGAGCGGAGCATGGTGG-3V. PCRs were performed in

200-Al tubes in a PTC-200 DNA Engine thermal cycler (MJ

Research, Inc., Watertown, MA, USA). Typically, a ‘‘touch-

down’’ PCR protocol was used. This consisted of an initial 5

to 10 cycles that included a denaturing step (92jC for 30 s),

an annealing step (starting at 72jC and going down to

66jC), and an extension step (72jC, 3 min). This was

followed by an additional 25 to 30 regular cycles (using

66jC as annealing temperature) and a 5-min final extension.

Acknowledgment

We thank P. Reed Larsen (Brigham and Women’s

Hospital, Harvard Medical School) for the human DIO3

cDNA clone.

References

[1] D.L. St. Germain, V.A. Galton, The deiodinase family of selenopro-

teins, Thyroid 7 (4) (1997) 655–668.

[2] V.A. Galton, E. Martinez, A. Hernandez, E.A. St. Germain, J.M.

Bates, D.L. St. Germain, Pregnant rat uterus expresses high levels

of the type 3 iodothyronine deiodinase, J. Clin. Invest. 103 (7) (1999)

979–987.

[3] J.M. Koopdonk-Kool, J.J. de Vijlder, G.J. Veenboer, C. Ris-Stalpers,

J.H. Kok, T. Vulsma, K. Boer, T.J. Visser, Type II and type III

deiodinase activity in human placenta as a function of gestational

age, J. Clin. Endocrinol. Metab. 81 (6) (1996) 2154–2158.

[4] E. Roti, L.E. Braverman, S. Fang, S. Alex, C.E. Emerson, Ontogen-

esis and placental inner ring thyroxine deiodinase and amniotic fluid

3,3V,5V-triiodothyronine concentration in the rat, Endocrinology 111

(1982) 959–963.

[5] T. Huang, I.J. Chopra, R. Boado, D.H. Solomon, G.N. Chua Teco,

Thyroxine inner ring monodeiodinating activity in fetal tissues of the

rat, Pediatr. Res. 23 (1988) 196–199.

[6] T. Huang, I.J. Chopra, A. Beredo, D.H. Solomon, G.N. Chua Teco,

Skin is an active site of inner ring monodeiodination of thyroxine to

3,3V,5V-triiodothyronine, Endocrinology 117 (1985) 2106–2113.

[7] T. Huang, A. Beredo, D.H. Solomon, I.J. Chopra, The inner ring (5-)

monodeiodination of thyroxine (T4) in cerebral cortex during fetal,

neonatal, and adult life, Metabolism 35 (1986) 272–277.

[8] A. Hernandez, M.J. Obregon, Presence of growth factors-induced

type III iodothyronine 5-deiodinase in cultured rat brown adipocytes,

Endocrinology 136 (10) (1995) 4543–4550.

[9] F. Courtin, P. Liva, J.M. Gavaret, D. Toru-Delbauffe, M. Pierre, In-

duction of 5-deiodinase activity in astroglial cells by 12-O-tetradeca-

noylphorbol 13-acetate and fibroblast growth factors, J. Neurochem.

56 (1991) 1107–1113.

[10] W. Croteau, S.L. Whittemore, M.J. Schneider, D.L. St. Germain,

Cloning and expression of a cDNA for a mammalian type III iodo-

thyronine deiodinase, J. Biol. Chem. 270 (1995) 16569–16575.

[11] D. Salvatore, S.C. Low, M. Berry, A.L. Maia, J.W. Harney, W.

Croteau, D.L. St. Germain, P.R. Larsen, Type 3 iodothyronine

deiodinase: cloning, in vitro expression, and functional analysis

of the placental selenoenzyme, J. Clin. Invest. 96 (5) (1995)

2421–2430.

[12] D.L. St. Germain, R. Schwartzman, W. Croteau, A. Kanamori, Z.

Wang, D.D. Brown, V.A. Galton, A thyroid hormone regulated gene

in Xenopus laevis encodes a type III iodothyronine 5-deiodinase,

Proc. Natl. Acad. Sci. USA 91 (1994) 7767–7771 (Erratum appears

in Proc. Natl. Acad. Sci. USA 91, 11282,1994).

[13] S. Van der Geyten, J.P. Sanders, E. Kaptein, V.M. Darras, E.R. Kuhn,

J.L. Leonard, T.J. Visser, Expression of chicken hepatic type I and

type III iodothyronine deiodinases during embryonic development,

Endocrinology 138 (12) (1997) 5144–5152.

[14] A. Hernandez, J.P. Park, G.J. Lyon, T.K. Mohandas, D.L. St. Germain,

Localization of the type 3 iodothyronine deiodinase (DIO3) gene to

human chromosome 14q32 and mouse chromosome 12F1, Genomics

53 (1) (1998) 119–121.

[15] A. Hernandez, G.J. Lyon, M.J. Schneider, D.L. St. Germain, Isolation

and characterization of the mouse gene for the type 3 iodothyronine

deiodinase, Endocrinology 140 (1) (1999) 124–130.

[16] H.M. Tu, G. Legradi, T. Bartha, D. Salvatore, R.M. Lechan, P.R.

Larsen, Regional expression of the type 3 iodothyronine deiodinase

messenger ribonucleic acid in the rat central nervous system and

its regulation by thyroid hormone, Endocrinology 140 (2) (1999)

784–790.

[17] C. Tsai, S.P. Lin, M. Ito, N. Takagi, S. Takada, A.C. Ferguson-Smith,

Genomic imprinting contributes to thyroid hormone metabolism in

the mouse embryo, Curr. Biol. 12 (2002) 1221–1226.

[18] A. Hernandez, S. Fiering, E. Martinez, V.A. Galton, D.L. St. Germain,

The gene locus encoding iodothyronine deiodinase type 3 (Dio3) is

imprinted in the fetus and expresses antisense transcripts, Endocrinol-

ogy 143 (11) (2002) 4483–4486.

[19] S.K. Nordeen, Luciferase reporter gene vectors for analysis of pro-

moters and enhancers, Biotechniques 6 (5) (1988) 454–457.

[20] A. Hernandez, D.L. St. Germain, Dexamethasone inhibits growth

factor-induced type 3 deiodinase activity and mRNA expression in


a cultured cell line derived from rat neonatal brown fat vascular–

stromal cells, Endocrinology 143 (7) (2002) 2652–2658.

[21] A. Hernandez, D.L. St. Germain, M.J. Obregon, Transcriptional acti-

vation of type III inner ring deiodinase by growth factors in cultured

rat brown adipocytes, Endocrinology 139 (2) (1998) 634–639.

[22] M. Kozak, At least six nucleotides preceding the AUG initiator codon

enhance translation in mammalian cells, J. Mol. Biol. 196 (1987)

947–950.

[23] W. Reik, J. Walter, Genomic imprinting: parental influence on the

genome, Nat. Rev. Genet. 2 (1) (2001) 21–32.

[24] W. Reik, J. Walter, Genomic imprinting: parental influence on the

genome, Nat. Rev. Genet. 2 (1) (2001) 21–32.

[25] C. Rougeulle, H. Glatt, M. Lalande, The Angelman syndrome candi-

date gene, UBE3A/E6-AP, is imprinted in brain, Nat. Genet. 17 (1)

(1997) 14–15.

[26] C. Rougeulle, C. Cardoso, M. Fontes, L. Colleaux, M. Lalande, An

imprinted antisense RNA overlaps UBE3A and a second maternally

expressed transcript, Nat. Genet. 19 (1) (1998) 15–16.

[27] J. Peters, S.F. Wroe, C.A. Wells, H.J. Miller, D. Bodle, C.V. Bee-

chey, C.M. Williamson, G. Kelsey, A cluster of oppositely im-

printed transcripts at the Gnas locus in the distal imprinting region

of mouse chromosome 2, Proc. Natl. Acad. Sci. USA 96 (7) (1999)

3830–3835.

[28] T. Li, T.H. Vu, Z.L. Zeng, B.T. Nguyen, B.E. Hayward, D.T. Bon-

thron, J.F. Hu, A.R. Hoffman, Tissue-specific expression of antisense

and sense transcripts at the imprinted Gnas locus, Genomics 69 (3)

(2000) 295–304.

[29] F. Sleutels, R. Zwart, D.P. Barlow, The non-coding Air RNA is re-

quired for silencing autosomal imprinted genes, Nature 415 (6873)

(2002) 810–813.

Complex organization and structure of sense and antisense transcripts expressed from the DIO3 gene...

Documents

Transcript of Complex organization and structure of sense and antisense transcripts expressed from the DIO3 gene...