Restriction enzymes were purchased from New England BioLabs and Toyobo. T4 DNA ligase and Agarose-LE Classic Type were supplied by TaKaRa. Klenow enzyme, T4 polynucleotide kinase and calf intestine phosphatase were obtained from New England BioLabs. AmpliTaq was from Roche. Oligonucleotides were purchased from Gene Design Inc. Phenylmethanesulfonylfluoride was provided by Sigma. MG115 and E-64d were from the Peptide Institute. All other chemicals used were of the highest grade commercially available.

pTA4-4ΔEXCTGm was constructed previously by cassette mutagenesis of pTA4-4ΔEX to introduce CTC or TTG codons instead of the CTG codons between the XmaI and BpU1102I sites [7] . The pTA4-4ΔEXCTGm was digested with XmaI, and an oligonucleotide cassette (Sense A/Antisense A) (Table 1) was inserted to regenerate the PEST sequence. The small XhoI-AccI fragment of the resulting plasmid, pTA4-4CTGm, was ligated

between the corresponding sites of pME-hGT1LΔEX to produce pME-hGT1LCTGm (Figure 1(a)). The small BglII-BclI fragment of pTA4-4CTGm was replaced with an oligonucleotide cassette [(Sense Ala1/Antisense Ala1) or (Sense Ala2/Antisense Ala2)] to produce pTA4-4CTGmAla1 or Ala2 (Figure 1(a)). The XhoI-AccI fragments of both plasmids were inserted into the corresponding sites of pME-hGT1LΔEX. The resulting plasmids were named pME-hGT1LCTGmAla1 and pME-hGT1LCTGmAla2, respectively. pTA4-4ΔEX and pME- hGT1LΔEX were constructed by Takeda et al. [7] and they were derivatives of pUC18 and pME18S, respectively.

To introduce further mutations into the first half of the PEST sequence (Figure 1(b)), the PvuI site between the SphI and NaeI sites of the vector sequence in pTA4-4CTGmAla2 was deleted by digestion with restriction enzymes SphI and NaeI, followed by Klenow enzyme treatment and self-ligation. The resulting plasmid, pTA4- 4dPCTGmAla2, was digested with BglII and PvuI. The PvuI-BglII fragment and the PvuI-PvuI fragment were ligated together with one of the following cassettes, STA, PA, PG, TPAA and SPAA (Figure 2 and Table 1). The XhoI-AccI fragments of the constructs (pTA4-4dPCTGmAla2STA, PA, PG, TPAA and SPAA) were inserted into the corresponding sites of pME-hGT1L. The resulting plasmids were named pME-hGT1LCTGmAla2STA, PA, PG, TPAA and SPAA, respectively (Figure 1(b)). The DNA sequence after each manipulation step was confirmed by the dideoxy chain-termination method [11] using sequence primers listed in our previous study [9] . The molecular biological techniques were performed by published methods [12] . Restriction enzyme digestion and ligation were carried out according to the manuals supplied by the manufacturers.

Cos-1 (ATCC) cells were grown in Dulbecco’s modified Eagle medium (GIBCO) supplemented with 7% (v/v) fetal bovine serum (JRH Biosciences). An expression plasmid was introduced into the cells by means of the diethylaminoethyl-dextran method as described previously [7] . Cells were grown in 5 ml of culture medium for 24 hours. Then, protease inhibitors (20 μM MG115, 1 mM phenylmethanesulfonyl fluoride and 50 μM E-64d) were added as a dimethyl sulfoxide solution (10, 25 and 25 μl/5ml medium, respectively). Cells were further grown for 24 hours.

A nuclear extract (10 μg protein) of transfected cells [7] was subjected to sodium dodecyl sulfate (SDS)-polya- crylamide gel-electrophoresis [13] , and then electro-blotted (200 mA, 90 min; ATTO Model-AE6675) onto an Immobilon^TM-P membrane [Millipore PVDF membrane (0.45 μm), IPVH00010]. The membrane was blocked overnight at 4˚C with phosphate-buffered saline [10 mM sodium phosphate buffer (pH 7.2), 137 mM NaCl, 3 mM KCl] containing 0.1% (v/v) Tween 20% and 3% (w/v) bovine serum albumin (Wako). Rabbit site-specific polyclonal antibody GATA-6N recognizing human S-type GATA-6 (Leu⁵⁹-Gln²¹⁷) [14] was used as the first antibody (1:500 dilution). Horseradish peroxidase-linked donkey anti-rabbit immunoglobulin (1:4000 dilution) (GE Healthcare) was used as the second antibody. Chemiluminescence was detected with a Western blotting kit (GE Healthcare) using Scientific Imaging Film (KODAK). Protein concentrations were determined with a BCA Protein Assay Kit (Pierce) using bovine serum albumin (Fraction V, Sigma) as a standard [15] .

The reporter gene assay was carried out as described by the published method [7] [9] . The GATA-6 expression plasmids used were pME-hGT1S (S-type), pME-hGT1L5’K (L-type), pME-hGT1L5’ΔEK (ΔE, L-type without Arg¹³-Gly¹⁰¹), pME-hGT1L5’ΔAK (ΔA, ΔE with the PEST sequence inserted), pME-hGT1L5’ΔA(-2)K (ΔA(-2), ΔE with the frame-shifted sequence inserted), and pME-hGT1L5’CTGmAlaK (ΔAla, ΔE with the Ala-cluster instead of the PEST sequence inserted). A Kozak sequence was introduced into the expression plasmids for various L-type GATA-6 proteins similar to in our previous study [16] .

pME-hGT1L5’CTGmAlaK was constructed as follows: pME-hGT1LCTGmAla1 was digested with EcoRV and AccI. pTA4-4dPCTGmAla2 was digested with PvuI and then treated with the Klenow enzyme. After AccI digestion, the small PvuI-AccI fragment was inserted into the large EcoRV-AccI fragment from pME- hGT1LCTGmAla1. The resulting plasmid (pME-hGT1LCTGmAla1,2) was subjected to PCR with primer pair MT006 and RN017, and then the amplified fragment was cloned into the TA vector. The NcoI-AccI fragment of the resulting plasmid and the XhoI-NcoI fragment from plasmid “MT005 - MT007 in TA” were excised, and then were ligated similarly in the case of pME-hGT1L5’ ΔK and pME-hGT1L5’ ΔA(-2)K (see details in Figure S3 in [16] ).

GATA-6 mRNA has an additional initiation codon located in frame upstream of the canonical one [7] , the latter being located at a similar position in the GATA proteins [4] . The nucleotide sequence around the upstream initiation codon does not match the Kozak sequence [17] , resulting in the translation of S-type GATA-6 from the downstream initiation codon in addition to L-type GATA-6 from upstream one [7] . It was further demonstrated that both species are not produced through proteolysis [9] , or that the CTG codons located in the intra-coding region of L-type GATA-6 mRNA does not function as translational initiation sites [7] . In this study, we used an expression plasmid of the L-type (pME-hGT1LCTGm) (Figure 2) as a control in which the CTG codons were substituted with CTC or TTG codons [7] to exclude any possibility that internal CTG codons could function as translational initiation sites [17] .

The PEST sequence (Glu³¹-Cys⁴⁶) located in the L-type specific sequence could be divided into two sub-se- quences; the amino terminal Pro-rich segment and the carboxyl terminal Ser-cluster (Figure 2 and Figure 3). We first examined the effect of the introduction of Ala residues into either of the entire sub-sequences (Ala1 and Ala2, respectively). The transient expression of such a mutant L-type GATA-6 in Cos-1 cells demonstrated that

the mobility of both proteins (Ala1 and Ala2) was increased on the gel compared with that of the control (CTGm) (Figure 3(a), lanes 1 - 4). However, the effect of substitution of the Pro-rich segment (Ala1) was much more significant; the increase of the mobility of Ala1 was 13%, while that of Ala2 was 8%.

We further studied the effects of substitutions in the amino terminal Pro-rich segment; S/T®A (Ser³³, Thr³⁴ and Ser³⁷ were substituted to Ala residues; also denoted to STA), P®A (Pro^{32, 35, 36, 38} were all substituted to Ala residues; also denoted to PA), P®G, (Pro^{32, 35, 36, 38} were all substituted to Gly residues; also denoted to PG), TP®AA (Thr³⁴ and Pro³⁵ were substituted to Ala residues; also denoted to TPAA) and SP®AA (Ser³⁷ and Pro³⁸ were substituted to Ala residues; also denoted to SPAA), in addition to the introduction of Ala residues into the Ser-cluster (Figure 2). The results showed that all the substitutions increased the mobility of the resulting recombinant proteins (Figure 3(a), lanes 5 - 9). The TP and SP substitutions in the Pro-rich sequence had similar medium effects to S/T®A. However, the substitution of all Pro residues (P®A and P®G) significantly increased the mobility (lanes 6 and 7), being similar to in the case of Ala1.

When we calculated the PEST score [10] (determined with genetic information processing software GENETYX^Ò MAC, Tokyo Japan) of the sequence “KPGDLPSTPPSPISSSSSSCDH” in the control CTGm sequence (Figure 2), a significantly high score, 14.78, was obtained. However, those for all mutagenized sequences in Figure 2 were negative values and were not defined as the PEST sequence. Thus, it seems likely that the control (CTGm) with the PEST sequence showed slow mobility and that the mobility was increased without the PEST sequence.

Next we determined the biological activity of L-type GATA-6 with or without the PEST sequence. For this purpose, the transcriptional activity of the reporter gene with the GATA-responsive promoter [9] was used together with the expression plasmid for GATA-6. To express only L-type GATA-6, a Kozak sequence was introduced ahead of the upstream initiation codon. As shown in Figure 4, the activity of the positive control with the PEST sequence (LΔA) was significantly higher. The activity of LΔA was as much as five times of that of LΔAla in which all the residues in the Pro-rich segment and the Ser-cluster were substituted with Ala residues. The activity of LΔAla was similar to those of LΔA(-2), LΔE and the S-type without the amino-terminal 146 residues of L-type GATA-6. These results suggest that the PEST sequence of L-type GATA-6 may play a functional role(s) in the transcriptional activation process other than as a constitutive degradation signal [10] . Furthermore, inhibitory sequence(s) for transcription may be present in the Arg¹³-Arg³⁰ and Arg⁴⁸-Pro¹⁰² regions within the amino-terminal extension of L-type specific 146 residues, since the L-type showed rather lower activity than LΔA.