Eur J. Biochcm 203,467-473 (1992) T; FEBS 1992

Functional study of the human thyroid peroxidase gene promoter Marc J. ABRAMOWICZ', Gilbert VASSART' and Daniel CHRISTOPHE'

' Institute of Interdisciplinary Research and ' Department or Medical Genetics, HGpital Erasme, Free University of Brussels, Belgium (Received September 13, 1991) - EJB 91 1224

Structure/function relationships in the human thyroid peroxidase gene promoter have been studied by deletion and mutation analyses and confronted with footprint patterns obtained with thyroid nuclear extracts and the purified thyroid transcription factor TTF-1. Crude nuclear extracts from dog thyroid primary cultures were shown to contain a binding activity recognizing the -119 to - 105 segment of the promoter (coordinates relative to the transcriptional start site). Deletion, or site-directed mutagenesis of this segment dramatically reduced transcriptional activity in transient expression experiments on gene fusions of the thyroid peroxidase promoter and the growth hormone reporter. This binding activity was increased in nuclear extracts from thyrocytes cultured in the presence of the CAMP-agonist forskolin. A mutation that decreased the promoter function in forskolin-stimulated thyrocytes resulted in weakening of the corresponding footprint. The binding site displays no significant sequence similarities with known CAMP-responsive elements. Mutagenesis of another region of the promoter (- 99 to - 94) induced the binding of an additional factor, resulting in a dramatically enhanced promoter activity. We show that the thyroid-specific transcriptional factor TTF-1 is not directly involved in the above-mentioned interactions and provide evidence suggesting that, in spite of displaying a similar binding pattern to thyroperoxidase and thyroglobulin promoters in vitro, TTF-1 plays a less important role in the former. Altogether, our data delineate the minimal thyroid peroxidase gene promoter in the human and identify the binding sites of two trans-activating factors, one of them being potentially the mediator of a non-conventional cAMP control, independent of the CAMP-responsive element and factor AP-2.

Thyroperoxidase (TPO) is the key enzyme for the synthesis of thyroid hormones [l]. It has been identified as the major component of the microsomal antigen involved in autoimmune thyroid diseases [2-41. Thyrotropin has been shown to increase both TPO enzymatic activity [5]and steady-state messenger RNA level in the thyrocyte of all species studied [6-81. This effect is mimicked by cAMP analogs and by forskolin, a universal activator of adenyiyl cyclase. It is still not known, however, whether TPO mRNA accumulation observed after stimulation by thyrotropin or forskolin is due to transcriptional or post-transcriptional mechanisms. According to recently published data [8, 91, no transcriptional regulation of TPO gene expression is observed in the immortalized, well differentiated rat thyroid FRTL-5 Correspondence lo M. J . Abramowicz, lnstitut de Recherche Intcrdisciplinaire, Campus Erasme, bgtiment C, 808 Routc de Lennik, B-1070 Bruxellcs, Belgium Ahbreviations. CRE, CAMP-responsive element; EGF, epidermal growth factor; PCR, polymerase chain reaction; TPO, thyroid peroxidase; TTF-1, thyroid transcription factor 1. Enzjwes. DNase (EC 3.1.21 . I ) ; restriction endonucleases SphI, BmmHI, IfindIII, Xbal, Kpnl and PstI (EC 3.1.21.4). Note. The novel nucleotide sequcnce data published here have been deposited with the EMBL sequence data bank and are available under the accession number X15103.

cell line. On the other hand, in dog thyrocytes in primary culture, a clear transcriptional regulation of the gene has been demonstrated in run-on assays [lo]. In contrast with the regulation of the thyroglobulin gene, activation of TPO gene transcription by CAMP is rapid and does not require on-going protein synthesis [ll]. The identification of a thyroid-specific enhancer upstream of the human TPO gene has recently been reported [12]. However, regulation of transcriptional activity by thyrotropin or CAMPwas not considered in this study. We have previously reported the isolation of the 5'flanking region of the human TPO gene and shown that a 900-bp segment confers thyrotropin and CAMP control of transcription to a reporter gene, when transfected in primocultured dog thyroid cells [13]. In the present study, using deletion and mutation analyses of constructs between genes for TPO and growth hormone, followed by DNase I protection experiments, we have delineated the functional portions of the TPO promoter and identified the binding sites for thyroid nuclear factors involved in trans-activations.

MATERIALS AND METHODS Plasmids construction All DNA manipulations used standard procedures [14J. The construct between genes for TPO and growth hormone

468 ( - 894hTPOGH) was formerly referred to as pTPOGH; its constructicelihas been described [I 31. The Sphi restriction site at - 519 (relative to the transcription start site) was used to generate a 0.5-kb fragment of the TPO gene promoter by Sphl and BamHI digestion of the - 894hTPOGH plasmid. The 0.5-kb fragment was subcloned into pBLCAT3 vector [15], from which it was further excised by Hind111 and BamHI digestions. It was then cloned into HindIII- and BamHI-digested pOGH plasmid vector [16] to yield the - 519TPOGH construct. The four other deleted TPO promoter/growth hormone fusions were obtained using the polymerase chain reaction (PCR) as described previously for the construction of the - 894hTPOGH clone. The same downstream primer complementary to the transcription start site region, bearing a BamHl restriction site, and an upstream primer complementary to the 5’ end of the deletion, containing a Hind111 restriction site, were used. The PCR products (extending to position + 14 relative to the TPO gene transcription start site), were digested by these enzymes and cloned into the HindIIIand BamHI-digested pOGH plasmid vector, yielding ( - 192hTPOGH, - 130hTPOGH, -94hTPOGH and - 67hTPOGH constructs, respectively (the coordinate relative to the cap site identifies the most 5’ undeleted base of the construct). The mutations were introduced in the - 192 deletion by trimolecular ligation of two PCR fragments into the pOGH vector using the following strategy: for each mutation, a couple of overlapping oligonucleotide primers oriented in opposite directions were synthesised ; the overlap corresponding to the segment of the TPO promoter to be mutated. The natural sequence was changed in such a manner as to create a restriction site (X), preferentially unique, in order to facilitate cloning and further screening of the recombinant clones. The two primers were used in two independent PCR reactions performed on the cloned genomic template, one with the upstream (HindIII) primer used to generate the - 192hTPOCH deletion, and the other with the downstream (BamHI) primer described above. The two PCR products were digested with the appropriate restriction enzymes, HindIII +X and BamHI X respectively, and ligated together into the HindIIIand RamHI-digested p0GH plasmid vector. The restriction sites chosen for the mutations (see Fig. 3A) were XbaI (constructs 132X, 126X and 78X), KpnI (constructs 112K, 103K and 94K), Pstl (construct 94P) and SphI (construct 84s). Sequence analysis revealed that construct 132X contained an additional 2-bp deletion, downstream from the tandemly mutated 6 bp. Construct 94DK‘ was obtained by blunting of the KpnI-digested 94K construct, using the exonuclease activity of the T4 DNA polymerase, which resulted in a 4-bp deletion. The sequence of each construct was verified on both strands by the dideoxy method [17].

+

metric assay of the cell culture medium collected at day 4 after transfection, respectively. The limit of detection of the growth hormone assay was 0.25 ng/ml. Preparation of nuclear extracts and DNaseI protection experiments Primary cultures of dog thyroid cells, prepared as described [18], were grown for three days in the presence of 1O0/o fetal calf serum + 25 ng/ml epidermal growth factor (EGF; referred to as ‘S’ conditions) or 10 pM forskolin (referred to as ‘F’ conditions). Nuclear extracts were prepared by a modification of a conventional procedure [22] as described in [23]. Briefly, cells were homogenized in 10 mM Hepes pH 7.9, 10 mM KC1, 1.5 mM MgCl,, 0.1 mM EGTA, 0.5 mM dithiothreitol, 0.5 mM phenylmethylsulfonyl fluoride; nuclei were pelleted in the presence of 6% sucrose, washed in the same buffer + 6.75% sucrose, resuspended in 10 mM Hepes pH 7.9, 400 mM NaCI, 1.5 mM MgCI,, 0.1 mM EGTA, 0.5 mM dithiothreitol, 0.5 mM phenylmethylsulfonyl fluoride, 5% glycerol; extracted nuclear proteins were precipitated by the slow addition of 0.33 g/ml (NH&S04. After centrifugation, the pellet was resuspended in 20 mM Hepes pH 7.9, 0.1 M KCI, 0.2 mM EGTA, 0.2 mM EDTA, 5% glycerol and dialysed against the same buffer with 20% glycerol. Protein concentration was determined by absorbance at 280 nm and aliquots were frozen in liquid nitrogen. These extracts, and the homeodomain portion of the recombinant TTF-1 protein [24] (a generous gift from R. Di Lauro), were used in DNase I protection experiments using 200-bp-long DNA probes derived from - 192hTPOGH, 94K and 103K constructs. These were end-labeled using [y3’P]ATP and T4 polynucleotide kinase at the Hind111 site at -192 (coding strand) or BamHI site at 14 (non-coding strand). The conditions for DNase I footprinting were as described [25]. Chemical sequencing of the end-labeled probes at the G + A residues [26] provided a reference sequence ladder.

+

RESULTS DNase I protection assays of the proximal human thyroid peroxidase promoter by thyroid nuclear extracts and the purified thyroid transcription factor TTF-1

Fig. 1 illustrates the protein - DNA interactions which can be demonstrated by DNase I protection on the 192-bp long human TPO promoter. The recombinant TTF-1 protein [24] (lane T) protects three main DNA segments in the human TPOproximal promoter at (- 168, - 154), (- 140, - 124) and (-77, -49) on the coding strand. By analogy with binding to the rat TPO promoter [27], these footprints will be referred to as A, B and C respectively. A lower-affinity binding site for TTF-1 might be present downstream from footprint C (about - 40 to - 30). Although somewhat fainter, these three Cell cultures, DNA transfection and reporter gene assays footprints are also observed with the crude thyroid nuclear Primary cultures of dog thyroid cells were prepared as extracts. N o difference was observed with extracts prepared described [IS]: 3-cm dishes were transfected by the di- from the cells cultured in serum + EGF (TPO gene not ethylaminoethyl-dextran method as described [19] using expressed) or forskolin (TPO gene expressed) conditions 1.6 pg test-plasmid DNA and 0.16 pg pSV,CAT plasmid [20] (compare lanes FI, Fh and S,, &). as an internal control of transfection efficiency. All constructs At - 119 to - 105, the thyroid nuclear extracts generate were tested in duplicate dishes in each transfection experiment. an additional footprint bordered by DNase-I hypersensitive Chloroamphenicol acetyltransferase and growth hormone sites (arrowheads, Fig. 1C). In contrast to the TTF-1 footassays were performed as described previously [13], by scintil- prints next to it, this footprint, together with associated hyperlation counting of tetramethylpentadecan/xylene-extracted sensitive sites, are clearly increased by forskolin treatment of butyrylated [3H]chloramphenicol [21] and immunoradio- the cells from which the extracts are made (lane F, versus S,

469

A

CODING

UPTSISIQF,.PU

B

NON-CODING

U P S t s 5 t T P U

A

d

+

“I:

-894

-519

0

t

-192

hTPO

-130

TTF-1 A

-94

-67

i i i

’

TTF-1

TTF-1 C

TATA

-500 -200

-900

+l

o

W’

Fig. 1. DNAse I protection assays o f the proximal TPO promoter. Nuclear extracts rrom dog thyroid cells cultured in conditions where the TPO gene is silent (serum - EGF, lanes S) or expressed (forskolin, lanes F) and recombinant TTF-1 protein (lane T) were used to protect a - 192 to + 14 TPO promoter fragment labeled terminally on the coding (A and C) or non-coding strand (B and C) from brief digestion by DNase I a t low (I, 3 mU/pI) or high (h, 6 mU/pl) concentrations. P, unprotectcd probe; U, G + A ladder. Coordinates are relative t o the transcriptional start ( + 1, broken arrow). CMR, CAMP-modulated binding. Arrowheads (C), DNasc I-hypersensitive sites

and Fh versus S,). Hence, it will be referred to as CAMPmodulated binding (CMB in figures). Nuclear extracts from dog thyrocytcs cultured in exactly the same conditions as the ‘control cells’ (i.e. cultured in the absence of any added agent) in the transient expression experiments (see below) gave results identical to the ‘S’ (serum-EGF) extracts (data not shown).

Fig. 2. Deletion analysis of the TPO promoter. (A) Extremities of the lruncations (vertical arrows) of the human TPO promoter (hTP0) with respect to the binding sites identified by footprinling (largcr rectangles) and putative regulatory elemcnts (smaller rectangles). C L , CRE-like motif; GC, G+C-rich motif. The coordinates are relative lo thc transcriptional start site (+ 1) represented by a broken arrow. ’ TATA, TATA box. (B) The various truncated constructs of the human TPO gene promoter (hTPO) fused to the growth hormonc reporter gene (GH) were transfected into primary cultures of dog thyroid cells; transcriptional activity was assayed by measuring the concentration of immunoreactive growth hormone in the culture medium. Columns and bars indicate the mean and range of duplicate transfection dishes in one representative transfection experiment. Values of growth hormone were normalized for transfection efficiency with a chloramphenicol acetyltransferase internal control (see Materials and Methods). The asterisks denote values below the detection limit

Table 1. Deletion analysis of the TPO promoter in cells treated with 10 pM forskolin. Values are means SEM oTn independcnt transient expression experiments, where thc transcriptional activities or the deleted promoter constructs (measured as in Fig. 2) were expressed as a percentage or that of the -894hTPOGH construct taken as ;I reference. pOGH, promoterless control.

Construct

Activity

Yo 100 214 59.9 32.2 2.8 2.9 2.5

- 894

hTPOGH - 51 9 hTPOGH - 192 hTPOGH - 130 hTPOGH 94hTPOGH - 67hTPOGH pOGH ~

Deletion analysis

The ability of promoters with 5‘ truncations to drive human growth hormone reporter gene expression was investigated in transiently transfected thyrocytes and compared with the parent -894hTPOGH construct [13]. Fig. 2A shows the 5‘ extremities of the truncations (arrows), with respect to the binding sites identified by footprinting and to putative binding sites found by inspection of the human TPO promoter sequence: in position -154 to -145, a DNA motif is found, TGACCAGTCA, which resembles the CAMP-responsive element (CRE) palindrome TGACGTCA [28] and is similar

f 21.8 ( n = k 4.8 (n = k 2.3 (n = k 0.37 (n = f 0.25 (n = 0.53 ( n =

4) 13) 12) 4) 3) 2)

to an element of sequence, TGACCAGCA, found in the promoter of the thyroglobulin gene (see Discussion). In position - 132 to - 123, a G C-rich decamerrepresents a potential binding site for factor AP-2 [29]. The results from one typical transfection experiment is shown in Fig. 2B. In agreement with the tight control of thc

+

470 A

lTF-1 A

2: ’

-065 TCCTAGCATCTT

CGGGCTATCCAAGCGCAGAGTCAGTTTAT~~GGTGGGTAACCAAGTCCCTI+AGG

C

B

10

5

A x

3aL

0 -192 wt

132 126

x

x

112 K

103

K

94 K

04

S

70 X

Fig. 3. Mutation analysis of the TPO promoter. (A) Mutations introduced in the 192-bp-long promoter segment of the -192hTPOGH construct. Each construct (132X, 126X, 112K, 103K, 94K, Y4P, 94DK’, 84s and 78X) harbors one of the underlined or overlined mutated windows, with the mutated sequence appearing in lower-case letters (bars denote deletions), along with the wild-type genomic sequence (in capital letters) with the TATA box and transcription start site double-underlined. The upper and lower open boxes represent the footprints (see Fig. I ) observed on the coding and non-coding strands, respectively. The putative AP-2 target is in faint letters. (B) Results of one representativc transfection cxperirnent of a primary culture of dog thyrocytes with the mutated constructs fused to the growth hormone (GH) reporter gcne. The transcriptional activity of each construct was assayed by mcasuring immunoreactive growth hormone accumulated in the culture medium. Columns and bars indicate the mean and range of duplicate transfcction dishes, respectively. Values of growth hormone were normalized for traiisfection efficiency with a chloramphenicol acetyltransfcrasc intcrnal control (see Materials and Methods). The asterisks denotc values below the detection limit. - 192wt, wild-type promoter. (C) Mutation analysis of the - 99 to -94 fragment of the TPO promoter in forskolin-stimulated cells. The effect of mutations Y4K, Y4P and 94DK’ were tested as described in B. Columns and bars indicate the mean and range of duplicate transfection dishes in one representative experiment

TPO gene transcription observed in cell cultures using run-on assays [I 11 and Northern blotting [30], none of the promoter constructs displayed detectable transcriptional activity in cells that wese not stimulated by thyrotropin or forskolin (i.e. in control medium). In forskolin-treated cells, a stronger promoter activity was reproducibly observed from the - 519 deletion as compared with the longer construct. Subsequent deletions resulted in weaker promoter activities (- 192 and - 130hTPOGH) that became insignificant for the two smaller constructs (- 94 and - 67hTPOGH), whose transcription activity was comparable to that of the promoterless pOGH control [13, 161. The results from several independent experiments are summarized in Table 1. Construct - 192hTPOGH contains all of the cis-elements described above. The TTF-1 site A, part of site B as well as the putative CRE-like element are deleted from construct - 130hTPOGH which displays about 30% of the activity of -894hTPOGH but keeps a tight control by forskolin. In contrast, removing the G + C-rich region and the CAMPmodulated binding site in the -94hTPOGH construct, virtually abolishes the transcriptional activity.

Except for the - 94 and the - 67hTPOGH constructs, the minimal estimate of the stimulation factor by forskolin was always greater than that observed with the metallothionein gene promoter in the pXGH5 control construct (13,161,which was 2.53 0.61 (mean f SEM from three independent experiments), indicating specific stimulation of the TPO promoter by forskolin. Mutation analysis Considering that the most dramatic drop In growth hormone expression occurred when removing the - 130 to -94 portion of the TPO promoter which contains the putative AP2 target (i. e the G +C-rich region) and the CAMP-modulated binding site, we further investigated the role of these sequences by introducing a series of mutations in the -192hTPOGH construct in a linker scanner approach. The resulting constructs harbored a partial or complete replacement of a window of 6 bp. Upon sequencing, construct 132X was shown to contain an additional 2-bp deletion downstream from the 6bp replacement mutation; the sequence of the mutations are shown in Fig. 3A, along with the corresponding wild-type sequence.

47 1 Tablc 2. Mutation analysis of the TPO promoter. Values are means f SEM of n independent transient expression experiments, where the transcriptional activities of the mutated promoter constructs (measured as in Fig. 3 B) were expressed as a percentage of the activity of the - lY2hTPOGH wild-type construct measured in cells treated with 30 pM forskolin. U, undetectable value. -60-

Construct

Activity of cells control

-70-

treated

-80-90-

-1Y2wt 132X 126X 112K 103K 94K 94P 94DK’ 84s 78X

U U U U U 37.4 f 10.2 (n = 5 ) U U U U

z 00 50.6 4.4 (n = 4) 233 f 32.6 (n = 4) 13.8 f 3.5 (n = 4) 15.7 2.1 ( n = 4) 1328 & 227 ( n = 6) 72.1 11.6(n = 3) 386 _+ 3.0 ( n = 3) 26.4 f 1.9 ( n = 4) 28.0 f 3.1 (n = 4)

3

--loo

-100-

1:

-110-

- -120

-120-

-140-

The results of one typical transient expression experiment of the mutated constructs is shown in Fig. 3 B, and the results from several independent transfection experiments are summarized in Table 2 . As with the deletion constructs, none of the replacement mutations except 94K (see below) exhibited detectable production of growth hormone when transfected in cells cultured under control conditions. In forskolin-stimulated cells, constructs 132X, 112K, 103K, 84s and 78X all exhibited significantly decreased promoter activity. Construct 132X, which disrupts the G +C-rich region, has only a slight inhibiting effect and the small effect of mutation 126X, which also involves the G +C-rich region, is stimulatory. In agreement with the deletion experiments, we observed a dramatic inhibitory effect of mutations 112K and 103K, both of which abolished 85% of the transcriptional activity of their wild-type - 192hTPOGH progenitor in forskolinstimulated cells. Together, these two mutations span the whole region (- 117 to - 103 relative to the transcriptional start site) involved in the interaction with the CAMP-modulated binding activity. Interestingly, the mutation in the 94K construct resulted in a very strong and reproducible stimulation of promoter activity, which was detected both in cells cultured in control medium and in forskolin-stimulated cells. In order to explore this phenomenon, two additional mutations were tested in the - 99 to - 94 region involved. The results are shown in Fig. 3 C and in Table 2 : while mutation 94P has little effect, if any, on promoter activity, mutation 94DK’ also strongly stimulates the CAMP-dependent transcriptional activity (see Discussion). The transcription-enhancing mutation in the 94K promoter construct was further investigated by footprinting experiments (see below). Mutations 84s and 78X both show a significant inhibitory effect, less pronounced, however, than that of 112K and 103K. In contrast to the 112K and 103K constructs, the level of reporter gene expression in forskolin-stimulated cells is here clearly compatible with specific stimulation of transcription by CAMP.

Fig. 4. Effects of TPO promoter mutations on nuclear factor binding. DNAse I protection experiments were performed on the mutated TPO promoter construct 94K, which shows increased transcriptional activity (A), and 103K, which shows decreased transcriptional activity (B), end-labeled on their coding strand ( - 192 to + 14 fragment) and protected with : lanes S, thyroid nuclear extract from serum-EGF conditions; lanes F, from forskolin conditions; lane T, recombinant TTF-I protein. P, unprotected probe; U, G A ladder. 1, low DNase I (3 mU/vl); h, high DNase 1(6 mU/pI). The locations of mutations are indicated by square brackets, Coordinates are relative to transcriptional start (+ 1, broken arrow). CMB, CAMP-modulated binding with DNase-I-hypersensitive sites (arrowheads). K94F, K94 Factor

+

Effects of TPO promoter mutations on nuclear factor binding Fig. 4 shows a DNase I protection experiment performed on two mutated promoters: the 94K (increased transcriptional activity) and 103K (decreased transcriptional activity). Nuclear extracts from thyrocytes cultured in serum-EGF (lanes S) or forskolin (lanes F) conditions were used to protect the mutated TPO promoter fragments (-192 to f14, coding strand labeled). Compared to the wild-type homologue (Fig. l), TTF-1 binding on both mutated constructs was unchanged (with crude thyroid nuclear extracts, lanes S and F, or the recombinant protein, lane T. On the 94K probe, an additional footprint centered on the 94K mutation was generated by the nuclear extracts. This footprint is not obtained with the TTF-1 recombinant protein (lane T). Upstream from the mutation, the CAMP-modulated footprint (CMB) was unchanged in the 94K construct. In contrast, this footprint was clearly diminished in the 103K construct at the level of the mutation, and the remaining binding activity was no longer increased by CAMP as it is in the wild-type promoter (Fig. 1) and 94K construct.

DISCUSSION We present here a first dissection of the human TPO gene promoter by structure/function analysis and nuclear factor

472 footprinting experiments using primocultured thyrocytes In conditions where expression is strictly dependent on CAMP. Previous work from our laboratory has focused on the expression of two thyroid-tissue-specific genes, the genes for TPO and thyroglobulin. In contrast to the latter, the transcription of the TPO gene in primary cultures of dog thyrocytes has been shown to be rapidly stimulated by CAMP,independently of on-going protein synthesis [ll]. A 900-bp fragment of the TPO gene 5’-flanking region was shown to mediate this effect 1131, which is specific as transcription from the metallothionein promoter (in pXGH5 construct [16])was only minimally affected by cAMP agonists. Genes subjected to rapid transcriptional control by cAMP usually contain a CRE in their proximal promoter region [31]; alternatively, the target site for factor AP-2 is sometimes found [29]. Interestingly, the 900-bp-long TPO promoter does not contain a CRE, but it displays a CRE-like motif which is furthermore similar to an element of the sequence found in the thyroglobulin proximal promoter [I91 (and our unpublished results). The deletion of this motif in construct -130hTPOGH does not abolish stimulation of the reporter gene transcription by CAMP.This demonstrates that the CRE-like motif is not required for transcription in response to cAMP stimulation nor for tissue-specific expression. The G + C-rich region at - 132 to - 123 was investigated because it represented a potential target for factor AP-2. Results from mutations 132X and 126X demonstrate that this fragment is also dispensable. The 50% reduction in promoter activity observed with construct 132X may be partly due to the modification in the DNA helix periodicity secondary to the 2-bp deletion (- 124 and - 123), which can have consequences on the interactions between trans-acting factors located at a distance from the deletion. Compatible with this observation, construct 126X, which mutates the 3’ half of the G + C-rich region without concomitant deletion, shows no decrease in CAMP-dependent transcriptional activity. If any, the effect of this mutation is slightly stimulatory. The TPO promoter region is thus conspicuous among the ‘type I’ CAMP-responsive genes [31] as it does not contain the CAMP-responsive cis-acting sequences known to date. The existence of a functional CRE equivalent with no similarity with canonical CREs has been reported in a short (82-bp) 5’flanking fragment of the growth hormone gene promoter [32]. The possible role of TTF-1, a recently cloned transcription factor important for the tissue-specific expression of the thyroglobulin gene, has been investigated. The implication of TTF-1 in the control of thyroglobulin gene transcription by cAMP remains an open question. As was shown for the rat TPO promoter [27], recombinant TTF-1 bound to three sites in the human TPO promoter. Although the consensus sequence for TTF-1 binding is highly degenerated, a significant similarity was found between rat and human sequences for the B and C sites but not for the A sites (Fig. 5). TTF-1 extracted from dog thyroid nuclei generated the same footprints and binding was the same whether the extract was prepared from thyrocytes not expressing (S) or expressing (F) the TPO gene. Although not conclusive, this result does not support the idea of TTF-1 mediating cAMP control. Deletion of site A ( - 130hTPOGH), partial deletion (id) or mutation (132X) of site B all weakened the promoter but did not abolish its activity, indicating a modest activation by TTF-1 at these levcls. Results from constructs 84s and 78X indicate a stimulatory function ofthe - 89 to - 74 region, where TTF-1 acting at site C might be one of the factors involved. Nonetheless, the absence of detectable activity of the -94 truncation indicates

. ..... . . ..... . .

HumTPO TTF-1 A : gatgacatggcactt-tgtttct RatTPO TTF-1 A :

aggtgccactcatagaaagc

............. . . ...................

HumTPO TTF-1 B : tcaggacacacaagaggcccggcgc RatTPO TTF-1 B : ccaggacacacaagca-cttg HumTPO TTF-1 C : gacgc-cactcga-cttcctagca

............. ...............

RatTPO TTF-1 C : gatgcccactcaagcttaga Fig. 5. Sequence comparison between TTF-1 sites in human and rat TPO promoters. Sequence similarity at site A is not significant

that, whether necessary or not,’ the integrity of site C is not sufficient for promoter function in TPO, which is in contrast to the thyroglobulin gene [23] (and our unpublished results). More promising as a possible target for factor(s) involved in regulation of transcription by cAMP is the - 119 to - 105 footprint. This segment of the TPO promoter is in a region whose deletion abolishes transcriptional activity in forskolinstimulated cells (Fig. 2); in contrast to TTF-1, the corresponding footprint displays a clear modulation by forskolin treatment of the thyrocytes used to prepare the extracts (Fig. 1). The functional importance of this binding site was confirmed by site-directed mutagenesis, with constructs 112K and 103K spanning the - 117 to - 103 segment, both of which dramatically reduced transcription. Moreover, the binding was shown to be decreased by a mutation (103K) decreasing the transcription of the reporter gene in forskolin-stimulated thyrocytes (Fig. 4). In the absence of a detectable promoter activity in cells cultured in control medium without forskolin, our transfection experiments with mutated constructs do not allow discriniination between cis-elements involved in CAMP-dependent regulation and those required to sustain basal promoter activity. However, the observed modulation by forskolin of CAMP-modulated binding activity footprinting at - 119, - 105 strongly suggests that it constitutes part of the cAMP regulatory machinery. Whether the footprinting reflects the binding of a true CAMP-activated transcription factor or that of a CAMP-induced tissue-specific assembly of factors remains to be determined. The latter possibility cannot be excluded in our primary cultures whose highly differentiated phenotype is strictly dependent on cAMP [33]. The 119, - 105 sequence does not display significant simihrity with any known regulatory element. Mutation in the - 99 to - 94 segment in the 94K construct strongly stimulated promoter activity (more than tenfold), and binding of a thyroid nuclear factor was shown at the level of the mutation. These results are best explained by the fortuitous creation of an enhancer site compatible with the architecture of the proximal promoter. Consistent with this view, construct 94P had little effect, if any, on promoter activity, and construct 94DK’, which contains a 4-bp inner deletion at the center of the 94K mutation, still possesses an enhanced (fourfold) activity despite a rupture of DNA helix periodicity and thus disalignment of trans-acting factors. This region of the TPO promoter seems to act as a hinge between the TATA box region and the upstream factors, whose mandatory role is demonstrated by loss of function produced by the - 94 truncation. Interestingly, the cAMP inducibility of the promoter has been conserved with all three 94 mutations. In this context. ~

473 growth hormone expression from construct 94K has proved useful for measuring intracellular variations in the concentration of CAMP secondary to transient expression of the A 2 adenosine receptor in cultured thyrocytes [34]. Also, study of the promoter activity of 94K-based mutated constructs which display detectable basal activity may prove useful in understanding CAMP regulation in our system. TTF-1 rccombinant protein was a generous gift from R. Di Lauro (EMBL, Heidelberg). We are grateful to Prof. J. E. Dumont and Dr M. Parmentier for useful discussions and continuous support and to Dr C. Gervy-Decoster for generous help. This work was supported by grants from the Fonds de lu Recherche Scientifique MCdicule, and the Belgian MinistPre de la Politique Scientifique. M. J. A. is a Research Assistant and D. C. a Research Associate at the Fond7 Nutianal de Irr Recherche Scientifique, Belgium. This text presents research results of the Belgian programme on Inter-University Poles of Attraction, initiated by the Belgian State, Prime Minister’s Office, Science Policy Programming. The scientific responsibility is assumed by its authors.

REFERENCES 1. Taurog, A. (1986) in The tl1,vroid (Ingbar, S. H. & Bravermann, L. E. eds) pp. 53-97, J. B. Lippincott Company, Philadelphia PA. 2. Czarnocka, B., Ruf, J., Ferrand, M., Canyon, P. & Lissitsky, S. (1985) FEBS Lett. 109, 147-152. 3. Portmann, L., Hamada, N., Heinrich, G. & De Groot, L. J . (1985) J. Clin. Endocrinol. Metuh. 61, 1001 - 1003. 4. Libert, F., Ruel, J., Ludgate, M., Swillens, S., Alexander, N., Vassart, G. & Dinsart, C. (1987) EMBO J . 6,4193-4196. 5. Magnusson, R. P. & Kapoport, B. (1985) Endocrinology 116, 1493- 1500. 6. Chazenbalk, G., Magnusson, R. P. & Rapoport, B. (1987) Mid. Endocrinol. 1, 91 3 - 9 17. 7. Nagayama, Y . , Yamashita, S., Hirayu, H., Izumi, M., Uga, T., Ishikawa, N., Ito, K . & Nagataki, S. (1989) J . Clin. Enclocrinol. Metuh. 68, 1155-1159. 8. Damante, G., Chazenbalk, G., Russo, D., Rapoport, B., Foti, D. & Filetti, S. (1989) Endocrinology 124, 2889-2894. 9. Foti, D., Gestautas, J. & Rapoport, B. (1990) Biochem. Biophys. Res. Commun. 168, 281 -287. 10. Gerard, C. M., Lefort, A,, Libert, F., Christophe, D., Dumont, J. E. & Vassarl, G. (1988) Mol. Cell. Endocrinol. 60, 239-242. 11, Gerard, C. M., Lefort, A,, Christophe, D., Libert, F., Van Sande, J . , Dumont, J. E. & Vassart, G. (1989) Mol. Endocrinol. 3, 21 10-21 18.

12. Kikkawa, F., Gonzalez, F. J. & Kimura, S. (1990) Mid. C‘c~ll.Bid. 10, 6216-6224. 13. Abramowicz, M. J., Vassart, G. & Christophe, D. (2990) B;ochm. Biophys. Re.s. Cornmun. 166, 1257- 1264. 14. Sambrook, J., Fritsch, E. 1;. & Maniatis, T. (1989) Moleculur cloning, u luhoratory manuul, Cold Spring Harbor Laboratory, Cold Spring Harbor NY 15. Luckow, B. & Schiitz, G. (1987) Nucloic Acids Rex. 15, 5490. 16. Selden, R. F., Howie, K. B., Rowe, M. E., Howard, M. G. & Moore, D. D. (1986) Mol. Cell. B i d . 6 , 3173-3179. 17. Sanger, F., Nicklen, S. & Coulson, A. R. (1977) Proc. Nut1 Accrd. Sci. U S A 74, 5463- 5467. 18. Roger, P. P. & Dumont, J. E. (1984) Mol. Cell. Endocrinol. 36, 79-93. 19. Christophe, D., C h a r d , C. M., Juvenal, G., Bacolla, A.. Teugels, E., Ledent, C., Christophe-Hobertus, C., Dumont, J . E. & Vassart, G . (1 989) Mid. Cell. Endocrinol. 64, 5 - 1 8. 20. Gorman, C . M., Moffat. L. F. & Howard, B. H. (1982) M i d . Cell. Bid. 2, 1044-1051. 21. Seed, B. & Sheen, J.-Y. (1988) Gene 67,271 -277. 22. Dignam, J . D., Lebovilz, R. M. & Roeder, R. G . (1983) Nucleic Acidr Res. 11, 1475-1489. 23. Civilareale, D., Lonigro, R., Sinclair, A . J. & Di Lauro, R. (1989) EMBO J . 8,2537 -2542. 24. Guazzi. S., Price, M., De Felice, M., Demante, G., Mattei, M.G . & Di Lauro, R. (1990) EMBO .I. 9,3631 -3639. 25. Musti, A. M., Ursini, V. M., Avvedimento, E. V., Zimarino, V. & Di Lauro, R. (1987) Nucleic Acid.s Rex. 15, 8149-8166. 26. Maxam, A. M. &Gilbert, W. (1980) Methods EnZy”KJ/.65.449525. 27. Francis-Lang, H., Price, M., Martin. U. & Di Lauro, R . (1990) Colhq. INSERM 207, 25 - 31. 28. Hoeffler, J. P., Meyer, T. E., Yun, Y., Jameson, J. L. & Habencr, J. F. (1988) Science 242, 1430- 1433. 29. Imagawa, M., Chiu, R. & Karin, M. (1987) Cell51, 251 -260. 30. Maenhaut, C., Brabant, G., Vassart, G. & Dumont, J. E. (1992). J . Biol. Chem. (in the press). 31. Roesler, W. J., Vandenbark, G. R. & Hanson, R . W. (1988) J . Biol. Chem. 263,9063 -9066. 32. Dana, S. &Karin, M. (1989) Mol. Endocrinol. 3, 815-821. 33. Roger, P. P., Van Heuverswyn, B., Lambert, C., Reuse, S., Vassart, G. & Dumont, J . E. (1985) Eur. .I. Biochem. 152,239 245. 34. Maenhaut, C . , Van Sande, J., Libert, F., Abramowicz, M., Parmentier, M., Vanderhaegen, J.-J., Dumont, J. E., Vassart, G. & Schiffman, S. (1990) Biochem. Biophys. Res. Commun. 173, 1169-1178.