Limitation in Use of Heterologous Reporter Genes for Gene Promoter Analysis

Various heterologous reporter genes have been widely used for the functional characterization of gene promoters. Many such studies often found weak to very strong silencer activities to be associated with specific parts of the basal promoter or further upstream regions. In this study, we carried out a systematic study on human blood coagulation factor IX (hFIX) and anti-coagulant protein C (hPC) genes, previously shown to have silencer activities associated with their 5′-flanking regions containing promoter sequences. With newly constructed chloramphenicol acetyltransferase (CAT) reporter vectors carrying hFIXor hPC gene promoter sequences, we confirmed the strong silencer activities associated with the regions nt −1895 through nt −416 of the hFIX gene or with the region nt −802 through nt −82 of the hPC gene. However, no such silencer activities associated with the specific regions were found when autologous hFIX cDNA, hFIX minigenes, orhPC minigenes were used as reporters in the expression vector system. Relative levels of CAT, hFIX, and hPC proteins produced in the transient assays correlated well with their mRNA levels. Human FIX minigene constructs containing a simian virus 40 (SV40) 3′-untranslated region (UTR) taken from the CAT reporter gene showed no silencer activity, indicating that SV40 3′-UTR sequence of the CAT reporter gene does not contribute to the silencer activity. Expression vectors constructed with the β-galactosidase gene under the control of hFIX gene promoter sequences also showed no silencer activity associated with the region nt −1895 through nt −416. These findings indicate that silencer activities associated with specific regions of promoter sequences as analyzed with CAT reporter genes may represent artifacts specific to the CAT reporter genes. Our findings strongly suggest a need for re-examination of promoter characterizations of many eukaryotic genes, which have been studied to date with CAT reporter genes.

Transcriptional regulation of eukaryotic genes are extremely diverse, strictly controlled and coordinated events and involve complex mechanisms for up-and down-regulation of gene expression (1,2). Such complexity and diversity of gene expression are obviously needed for maintaining intricate balances among biological reactions and systems in response to various internal and external stimulations and stress. For characterization of promoter functions of various genes, heterologous reporter genes, such as CAT, ␤-galactosidase, luciferase, and green fluorescent protein genes, have often been utilized (3)(4)(5). Use of such heterologous reporter genes often gives convenience in both qualitative and quantitative analyses of promoter activities. At the same time, however, use of such heterologous reporter genes may result in significant biases in assessing subtle structure-function relationships of promoters and irrelevant observations, which are primarily due to the introduction of foreign elements or due to the elimination of intrinsic elements of test genes in the assay system. Any regulatory machinery including those in the promoter regions must have evolved in the context of the rest of gene structures, neighboring genomic elements as well as of higher order structures of chromatin and chromosome. Therefore, use of a heterologous reporter gene in studying regulatory mechanisms of an unrelated gene may have a risk of resulting in irrelevant observations and conclusions. To date, a large number of mammalian genes have been analyzed for their promoter functions with various heterologous reporter genes. However, literally no systematic studies have been conducted addressing the issue concerning the relevancy and limitation of commonly used heterologous reporter genes.
Many functional analyses of promoters of eukaryotic genes to date by using CAT 1 reporter genes have shown weak to strong negative regulatory activities or silencer activities, which are associated with specific 5Ј-flanking regions containing promoter sequences (6,7). Human genes for FIX and PC, important factors involved in blood coagulation and anti-coagulation pathways, respectively (8,9), are among them (10,11). The 5Ј-upstream region of the hFIX gene promoter beyond approximately nt Ϫ800 up to at least nt Ϫ1900 showed a very strong silencer activity in the CAT vector context (10). The hPC gene promoter was also shown to have a substantial silencer activity associated with the 5Ј-upstream region beyond nt Ϫ82 up to nt Ϫ802 (11). Literally nothing was known about the mechanisms by which these silencer activities are generated and what roles they play in the natural regulation of hFIX and hPC gene expression.
In this report, using transient expression assay system with HepG2 cell or HTC cell lines, we first demonstrate that CAT reporter gene constructs reproducibly show silencer activities associated with hFIX and hPC gene promoters. We then present experimental evidence that such silencer activities are irrelevant artifacts specifically associated with the CAT reporter gene, but not with other reporter genes including autologous hFIX or hPC gene as well as the ␤-galactosidase gene, another commonly used heterologous reporter gene.

EXPERIMENTAL PROCEDURES
Materials-Restriction enzymes and DNA modification enzymes were purchased from Invitrogen and New England Biolabs. Radioactive nucleotides, [␣-32 P]dCTP, were obtained from Amersham Biosciences. Mouse anti-hPC monoclonal antibody and rabbit polyclonal anti-hPC antibodies were purchased from Celsus Laboratory. Horseradish peroxidase-linked goat anti-rabbit IgG was purchased from Invitrogen. Anti-hFIX polyclonal and monoclonal antibodies for enzyme-linked immunosorbent assay (ELISA) were obtained from Hematologic Technologies Inc and Enzyme Research Laboratory, respectively, as previously described (12). Media, fetal calf serum, penicillin, and streptomycin for mammalian cell cultures were obtained from Invitrogen. FuGENE-6 transfection reagent and CAT ELISA kit were purchased from Roche Molecular Biochemicals. HepG2 cells were obtained from ATCC. Rat hepatoma cells, HTC cells, was kindly provided by Dr. Thomas Gelerhter in this department. Human PC vector pUC119-hPC was kindly provided by Dr. Francis Castelino at the University of Notre Dame. o-nitrophenyl ␤-D-galactoside for the ␤-gal assay was purchased from Sigma. All other reagents were of the highest quality commercially available.
Construction of CAT and ␤-Gal Expression Vectors-Plasmids pCH110 and pSV2CAT contain ␤-gal and CAT genes under the control of the SV40 early promoter, respectively (10,13). Plasmid pUMSVO-CAT, a promoter-less CAT vector with virtually no background CAT expression activity, was used as a control CAT expression vector (14).
Construction of CAT expression vectors under controls of the hFIX or hPC promoters are as follows: expression vectors with hFIX 5Ј-flanking sequences, Ϫ1895FIX/CAT and Ϫ416FIX/CAT, were constructed by inserting hFIX gene fragments spanning nt Ϫ1895 through nt ϩ29 or nt Ϫ416 through nt ϩ29, which were PCR-amplified from minigene Ϫ2231FIXm1 (15), into pUMSVOCAT at the SmaI site, respectively (Table I). Expression vectors, Ϫ802PC/CAT and Ϫ82PC/CAT, were constructed by inserting PCR-amplified fragments with SmaI linkers, which span the region nt Ϫ802 through nt ϩ66 or nt Ϫ82 through nt ϩ66 of the hPC gene, into pUMSVOCAT, at the SmaI site, respectively (11,14) (Table I). A lacZ gene fragment with BamHI and HindIII sticky ends (3736 bp in size) were released from pCH110 (10,16), followed by their insertion into pUC19 between BamHI and HindIII sites (3), thus generating pOGAL. A PCR-amplified fragment of 1924 bp (nt Ϫ1895 through nt ϩ29 of the hFIX gene) or 445 bp (nt Ϫ416 through nt ϩ29) with HindIII sticky end was then inserted into pOGAL at HindIII site, thus generating Ϫ1895FIX/␤GAL or Ϫ416FIX/␤GAL. All the PCR-amplified sequences and ligation site sequences of expression vectors were subjected to automated dideoxy sequencing to confirm their accuracy.
Construction of hFIX and hPC Expression Vectors-Human FIX minigene (hFIXm1) and cDNA (hFIXc) expression vectors, Ϫ1895FIXm1, Ϫ416FIXm1, Ϫ1895FIXc, and Ϫ416FIXc, were constructed with plasmid pUC19. These vectors contained 5Ј-end regulatory regions of the hFIX gene identical to those used in the CAT vectors. Constructs Ϫ416FIXm1 and Ϫ416FIXc were prepared as previously described (12). Constructs Ϫ416FIXm1 or Ϫ416FIXc were digested with SphI and StuI, removing a 483-bp fragment encompassing a region nt Ϫ416 through nt ϩ67 of the hFIX gene (the unique StuI site at nt ϩ67 in exon I). A PCR-amplified 1967-bp fragment (nt Ϫ1895 through nt ϩ72 of the hFIX gene) with SphI and StuI sticky ends at 5Ј-and 3Ј-ends, respectively (Table I), was then inserted at the sites, thus producing Ϫ1895FIXm1 and Ϫ1895FIXc, respectively. Minigenes Ϫ1895FIXm1 and Ϫ416FIXm1 were digested by BamHI and KpnI, removing the 3Ј-UTR and poly(A) signal sequence of the hFIX gene, followed by treatment with Klenow enzyme. A fragment containing the SV40 early region and poly(A) signal sequence (135 bp in size) was released by HpaI and BamHI double digestion of pUMSVOCAT, followed by treatment with Klenow enzyme, and then inserted into the above fragment vectors, generating Ϫ1895FIX/SV40 and Ϫ416FIX/SV40, respectively. Minigene Ϫ416FIXm1 was digested by SphI/NheI to remove its hFIX promoter region, followed by Klenow enzyme treatment and self-ligation, thus generating pFIXm1, a promoter-less hFIX minigene control vector for hFIX expression assays. A promoter-less hFIX cDNA control vector, pFIXc, was similarly generated by SphI/NheI double digestion of Ϫ416FIXc, removing the promoter region.
Human PC minigene (hPCm1) expression vectors, Ϫ802PCm1 and Ϫ82PCm1, were constructed as follows: a fragment spanning nt Ϫ802 through nt ϩ1560 of the hPC gene (2362 bp in size) with SphI and MscI linkers at 5Ј-and 3Ј-ends, respectively, was PCR-amplified by using human genomic DNA as a template (11,17) (Table I), and inserted into pUC119-hPC between SphI and MscI sites to replace its 5Ј-end portion. The 3Ј-end portion of the resultant construct containing the internal Sse8387I site in the 3Ј-UTR through an EcoRI site present at the 3Ј-end immediately outside of the poly(A) attachment site, was released by Sse8387I/EcoRI double digestion. A PCR-amplified fragment (615 bp in size, spanning nt ϩ10494 through nt ϩ11108 of the hPC gene) with Sse8387I and EcoRI sticky ends (17), was then inserted to fill the gap, thus generating Ϫ802PCm1. Construct Ϫ82PCm1 was similarly generated by replacing the 5Ј portion of Ϫ802PCm1 (nt Ϫ802 to nt ϩ1560) with a fragment spanning nt Ϫ82 to nt ϩ1560. The 5Ј-end region, non-coding exon 1 and partial intron 1 of the hPC gene were removed from Ϫ802PCm1 by complete and partial digestion with SphI and MscI, respectively. The remaining fragment from Ϫ802PCm1 was treated by Klenow fragment, and then self-ligated, generating promoter-less control vector, pPCm1. Expression vectors, Ϫ802PC/FIXc and Ϫ82PC/FIXc, were constructed as follows: a fragment spanning nt Ϫ802 through nt ϩ66 of the hPC gene with SphI and NheI linkers at the 5Ј-and 3Ј-ends, respectively, was PCR-amplified from Ϫ802PCm1 (Table I). After digestion with SphI/NheI, this fragment was inserted into Ϫ416FIXc, which was digested in advance with SphI/NheI, removing the hFIX promoter sequence, thus generating Ϫ802PC/FIXc. Construct Ϫ82PC/FIXc was similarly generated by replacing the 5Ј-end fragment (nt Ϫ802 through nt ϩ66) of Ϫ802PC/FIXc with a 148-bp hPC fragment spanning nt Ϫ82 through nt ϩ66 of the hPC gene. All the PCR-amplified sequences and ligation site sequences of newly constructed expression vectors were confirmed by automated dideoxy sequencing for their accuracy.
a The locations of primers are shown by nucleotide residue numbers in the parentheses.
Cell Culture and Transfection-HepG2 cells were cultured in Dulbecco's modified Eagle's medium (DMEM) supplemented with L-glutamine, 25 mM HEPES buffer, 110 mg/liter sodium pyruvate, antibiotics (penicillin, streptomycin, and neomycin), and 10% fetal bovine serum in a 5% CO 2 atmosphere at 37°C. HTC cells were cultured under similar conditions except 5% fetal bovine serum supplemented. Cell transfection was carried out with the FuGENE 6 transfection reagent as previously described (18). Plasmid pCH110 was used as a transfection internal control for CAT, hFIX and hPC expression vectors, while pSV2CAT was similarly used for ␤-gal expression vectors. 4 -5 independent assays were carried out, and average activities were shown with S.D.
CAT and ␤-Gal Assays-The CAT protein assay was carried out by using a CAT ELISA kit according to the manufacturer's instruction. ␤-gal activity was assayed as previously reported (19). Cell extracts obtained from cells co-transfected with pCH110 or pSV2CAT were first assayed for ␤-gal or CAT activities, respectively. These activities were used for normalizing transfection efficiencies among culture dishes. Amounts of cell extracts taken for activity assays were adjusted for the optimal assay range of CAT or ␤-gal activity.
Human FIX and hPC ELISA Assay-Human FIX produced into the conditioned culture medium was quantified by hFIX-specific ELISA as previously described (15). This ELISA system reproducibly detected hFIX antigen as low as 1 ng/ml. Human PC produced was assayed by ELISA as previously described (20). Horseradish peroxidase-conjugated goat anti-rabbit IgG was used as the detection antibody. For each expression vector, minimally three independent ELISA were carried out with duplicated assays for each diluted culture medium sample, and average values were calculated to determine the amounts of protein produced.
Northern Blot Analysis of hFIX, hPC, and CAT mRNA Levels in the Transfected Cells-Northern blot analysis of HepG2 cells transfected with hFIX, hPC, or CAT constructs was carried out as previously described (15,20). Total RNA samples prepared from the transfected HepG2 cells were subjected to agarose gel electrophoresis (15 g per lane). For CAT mRNA detection, a coding region fragment of CAT gene (624 bp in size) was PCR-amplified with 5Ј and 3Ј primers (5Ј-ACCAC-CGTTGATATATCC-3Ј and 5Ј-CTGCCACTCATCGCAGTA-3Ј, respectively) by using pUMSVOCAT as a template, and was used as a hybridization probe. A fragment (588 bp in size) prepared by SspI/BamHI digestion of Ϫ416FIXm1 was used for hFIX hybridization. Human PC hybridization probe (365 bp in size, from nt ϩ8385 to ϩ8749 in genomic nucleotide numbering) was prepared as previously described (20). These probes were labeled with [␣-32 P]dCTP by random priming (Amersham Biosciences) to a specific activity of ϳ1 ϫ 10 9 cpm/g. To confirm the presence of equal amount of RNAs, blotted filters were washed twice at 75°C, each for 30 min, in 10% sodium dodecyl sulfate, and hybridized with labeled RNR 18 probe.

RESULTS
Transient Expression Activities of CAT, hFIX Minigenes, and cDNA Expression Vectors-Transient expression activities of newly generated CAT reporter gene constructs with different hFIX 5Ј promoter sequences, Ϫ1895FIX/CAT and Ϫ416FIX/ CAT, in HepG2 cells are shown in Fig. 1A. Construct pUMS-VOCAT, a promoter-less control, gave no detectable CAT activities. Construct Ϫ1895FIX/CAT containing the hFIX promoter region nt Ϫ1895 through nt ϩ29 showed only 15.9% CAT activity of that of the Ϫ416FIX/CAT containing the region nt Ϫ416 through nt ϩ29. In agreement, Northern blot analyses showed a similarly lowered mRNA level in HepG2 cells transfected with Ϫ1895FIX/CAT, ϳ20% of that of cells transfected with Ϫ416FIX/CAT (Fig. 1D). These results indicated the existence of a strong silencer activity associated with the 5Јupstream region of the hFIX gene, confirming our previous observation (10).
Relative expression activities of hFIXm1 constructs containing the hFIX promoter sequences identical to those used in CAT constructs, Ϫ1895FIXm1, Ϫ416FIXm1, Ϫ1895FIXc, and Ϫ416FIXc, are shown in Fig. 1, B and C. Construct Ϫ416FIXm1 expressed hFIX at a level of ϳ40 ng/10 6 cells per 24 h and was defined to be at 100% activity. HepG2 cells expressed non-detectable levels of hFIX as previously reported (15). As expected, the promoter-less control vector, pFIXm1, showed only minimal hFIX expression activity. Both Ϫ1895FIXm1 and Ϫ416FIXm1 showed equivalent transient expression activities, which were correlated well with similar hFIX mRNA levels in HepG2 cells transfected either with Ϫ1895FIXm1 or with Ϫ416FIXm1 (Fig. 1E). These results indicated no appreciable silencer activities associated with the region nt Ϫ1895 through nt Ϫ416 of the hFIX gene. Similar to the hFIX minigene constructs, expression vectors with hFIX cDNA sequence, Ϫ416FIXc and Ϫ1895FIXc, showed equivalent hFIX expression activities in HepG2 cells, confirming no silencer activities associated with the region nt Ϫ1895 through nt Ϫ416 (Fig. 1C). This result also suggested that there is no specific contribution of the functional hFIX intron 1 sequence to the silencer activity shown by CAT reporter analysis.
Transient Expression Activities of CAT and hPC Minigenes Expression Vectors-Transient expression activities of newly constructed CAT reporter gene vectors with hPC promoter sequences, Ϫ802PC/CAT and Ϫ82PC/CAT, are shown in Fig.  2A. The average CAT transient expression activity of Ϫ802PC/ CAT was only 21.6% of that of Ϫ82PC/CAT, and in agreement, the CAT mRNA level of HepG2 cells transfected with Ϫ802PC/ CAT was also lowered to ϳ25% of Ϫ82PC/CAT (Fig. 2C). These results indicated the presence of substantial silencer activity associated with the 5Ј-upstream region (nt Ϫ82 to nt Ϫ802) of the hPC gene, which is consistent with the previous report by Miao et al. (11).
Transient expression activities of hPC minigenes expression vectors were assayed with both HepG2 cells and HTC cells, respectively. Relative expression activities of hPC miningene constructs with that of Ϫ82PCm1 defined as 100% are shown in Fig. 2B. HepG2 cells showed a low, but significant level of endogenous hPC expression (ϳ12 ng/10 6 cells per 24 h, data not shown), which correlated with the significant background expression activity of pPCm1, a promoter-less construct (average 12.8 ng/10 6 cells per 24 h). Construct Ϫ82hPCm1 produced 101 ng/10 6 cells per 24 h into the culture medium. Constructs Ϫ82PCm1 and Ϫ802PCm1 gave equivalent hPC expression activities in HepG2 cells, correlating well with the similar levels of hPC mRNA in the transfected HepG2 cells (Fig. 2D). Relative expression activities of hPC expression vectors assayed with HTC cells, a rat hepatoma cell line with no endogenous hPC expression, agreed well with those obtained in HepG2 cells (Fig. 2B). These results obtained with native gene constructs contradict the observation with CAT reporter gene ( Fig. 2A).
Possible Effects of the SV40-derived 3Ј-UTR on the Expression Activities of hFIX Expression Vectors-Relative expression activities of chimeric pFIXm1/SV40, Ϫ1895FIXm1/SV40 and Ϫ416FIXm1/SV40 vectors, in which SV40 3Ј-UTR including poly(A) signal sequence replaced the counterpart of the hFIXm1 gene, are shown in Fig. 3A. Both constructs showed

Transient Expression Activities of ␤-Gal Expression Vectors with hFIX Promoters and hFIX cDNA Expression Vectors with
hPC Promoters-Transient expression of ␤-gal reporter gene constructs under the control of the hFIX gene promoter sequences, Ϫ1895FIX/␤GAL and Ϫ416FIX/␤GAL, were assayed with HepG2 cells (Fig. 3B). Construct pO␤Gal, a promoter-less control, gave no appreciable ␤-gal expression in HepG2 cells. No silencer activity was found to be associated with the region nt Ϫ1895 through nt Ϫ419 of the hFIX gene in the ␤-gal reporter gene context. Constructs composed of hPC promoter sequences and hFIX cDNA reporter, Ϫ802PC/FIXc or Ϫ82PC/ FIXc, also showed no silencer activity associated with the 5Јflanking region nt Ϫ802 through nt Ϫ82 of the hPC gene (Fig.  3C). DISCUSSION Heterologous reporter genes have been widely used for functional characterization of many eukaryotic gene promoters. Use of such reporter genes, however, may have a risk of generating irrelevant results and conclusions on gene regulation. This is the issue that we systematically addressed in the present study. Such irrelevant results may be generated due to the absence of structural elements in the heterologous reporter gene, which are present in the native test gene and required for its regulation in the context of its own promoter. Alternatively, structural elements in the heterologous genes, which are absent in the test genes, may critically affect the transcriptional regulation of test genes.
In the present studies, we first demonstrated that CAT reporter gene expression vectors reproducibly showed strong silencer activities associated with the 5Ј-flanking regions of the hFIX gene (Fig. 1A), confirming the previous findings (10). Such silencer activities, apparently due to transcriptional suppression as demonstrated by the similarly reduced mRNA levels (Fig. 1D), however, were not found when hFIX minigene or cDNA were used as a reporter in place of the CAT gene (Fig. 1,  B and C). These results indicated that the silencer activities found to be associated with the 5Ј-specific upstream region of the hFIX gene may be irrelevant to the hFIX gene, and were generated through a specific combination of the hFIX promoter and the CAT reporter gene. This hypothesis was further supported by the similar findings obtained with hPC gene expression vectors (Fig. 2). Use of the CAT reporter gene in combination with the hPC promoter region gave substantial silencer activity associated with the 5Ј-flanking region nt Ϫ802 through nt Ϫ82 (Fig. 2A), whereas no such silencer activity was observed when the hPC minigenes were used as a reporter (Fig.  2B).
It is possible that some elements present in the hFIX minigene, such as the functional intron sequence which grossly elevates mRNA levels (12,21), might eliminate the silencer activity observed with CAT reporter. However, this possibility was not supported, since hFIX cDNA constructs without the intron sequence did not show any silencer activities (Fig. 1C). The SV40 3Ј-UTR sequence used in the CAT reporter gene was also shown to be not responsible for generating the silencer activity (Fig. 3A). Furthermore, the observed correlation between levels of produced CAT, hFIX, or hPC protein and their mRNA levels eliminates the possibility of reporter gene-dependent differences in translational efficiency, intracellular protein trafficking or secretion process as a possible cause for generating CAT reporter gene-specific silencer activities (Figs.  1, D and E and 2, C and D). Examinations with another series of chimera constructs composed of promoter sequences of the hPC gene connected with the hFIX cDNA sequence also showed no silencer activity associated with the hPC promoter region nt Ϫ802 through nt Ϫ82 (Fig. 3C). It is important to emphasize that hFIX and hPC genes share a significant similarity in their coding regions (22). However, their promoter regions are grossly different from each other, suggesting that these promoter regions took different evolutional pathways (20,23).
Expression vectors with the ␤-gal gene, another common reporter gene, also do not show any silencer activity with hFIX promoter sequences, further supporting our conclusion that the silencer activities observed with the 5Ј-flanking upstream-specific regions of hFIX as well as the hPC gene are unique artifacts generated by use of CAT reporter genes (Fig. 3B). The detailed mechanisms responsible for generating such silencer activities associated with the CAT reporter genes in combination with the promoters of genes of interest remain to be determined. Particularly, identification of any specific parts of the CAT gene structure involved in cross-talk with the specific regions of test promoter sequences would be critical.
Other heterologous reporter genes, which are not included in the present study, may also exert unusual effects including silencer activities on various gene promoters. Furthermore, specific combinations of a tested gene promoter and a heterologous reporter gene may even generate pseudo-high activity for the promoter although to our knowledge no systematic studies are reported to date. These possibilities are yet to be tested.
Our findings indicate that observations on the promoter structure-function relationship analyzed with the commonly used CAT reporter gene may not necessarily represent the true promoter regulatory mechanisms of many genes of interest, and strongly suggest a need for their systematic re-examination. Analysis of a specific promoter may be best done, if possible, with its autologous coding and subsequent downstream structure sequences as a reporter gene.