Mutagenesis by reversible promoter insertion to study the activation of NF-κB

Results

A Promoter Is Required for Retroviruses to Increase Mutagenesis. Insertion of a promoter in or near a gene may perturb its function, generating a product that acts in a dominant manner. We predicted that this phenomenon, rather than the insertion of a DNA fragment per se, would be the major mechanism of retroviral mutagenesis. Alternatively, the insertion of a DNA fragment that disrupts a target gene might lead to a selectable phenotype either through a reduction in gene dosage alone or in concert with sporadic loss of the untargeted allele. We tested the mechanism of mutagenesis by using a genetic system previously established in our laboratory (6). HCT9 cells, derived from HT1080 fibrosarcoma cells, were engineered to express the genes for the thymidine kinase from herpes simplex virus and puromycin N-acetyl transferase, both under the control of p53-responsive promoters. The cells were also modified to express a murine ecotropic receptor that makes them highly susceptible to infection with ecotropic retroviruses. HCT9 cells have intrinsically high levels of p53 and, therefore, are constitutively resistant to puromycin and sensitive to GCV. Mutants with a reversed pattern of drug resistance can be generated, for example, by chemical mutagenesis, and are likely to reflect events that abrogate p53-dependent transactivation.

We infected HCT9 cells with a pBNGFP retrovirus that carries Moloney murine leukemia virus LTRs with a functional promoter (Fig. 1A). In parallel, we used a derivative of this vector in which the promoter in the 3′ LTR is inactivated by deletion, creating a self-inactivating construct. After transfection into packaging cells, both constructs generate similar titers of infectious virus, as determined by the incidence of G418 resistance in independently infected 293 cells (data not shown). However, upon reverse transcription and integration, only pBNGFP retains functional promoters. When infected cells were plated in the presence of GCV, we observed that retention of a functional LTR increased colony formation by more than an order of magnitude, whereas the number of mutants generated by the promoterless construct was indistinguishable from the number found in mock-infected cells (Fig. 1B).

We conclude that, in our system, promoter insertion is the major determinant of retroviral mutagenesis, exceeding by far the contribution of the insertion of inert DNA. The functional consequences of promoter insertion probably reflect the overexpression of a diffusible product, and therefore, such a mutation is dominant. In these mutants, the selective pressure to lose the intact allele should be minimal, and therefore, inactivation of the inserted promoter, even without complete restoration of the targeted allele, is likely to result in phenotypic reversion due to the presence of the intact allele.

Improvement of Insertional Mutagens. Southern analysis of the mutants generated by infection of HCT9 cells with pBNGFP revealed that five to seven integration events were present in each clone examined (data not shown). A high multiplicity of infection increases the yield of insertional mutants without increasing the frequency of spontaneous mutations: the former is a function of total number of inserts, whereas the latter should be approximately constant. Therefore, with multiple inserts in each cell, the same genetic experiment can be conducted with fewer cells, and fewer false-positive clones would have to be analyzed. However, these benefits are balanced by the need to conduct full biological studies on each of the integration sites. The use of a regulatable promoter does not offer a solution to this problem because all of the inserts are likely to be regulated similarly. To circumvent this limitation, we used site-specific recombination to physically remove the functional promoter. Interestingly, the efficiency of site-specific recombination seldom reaches 100% (e.g., ref. 15). For a given mutant clone, imperfect efficiency of excision should generate progeny that differ in the numbers and patterns of inserts remaining, and a specific insert could be correlated with the mutant phenotype.

We generated a series of retroviral vectors optimized for insertional mutagenesis (Fig. 2 A and B) by using the Moloney murine leukemia virus backbone of the pBabe vectors (9). We positioned the recognition sequence for Cre recombinase inside the 3′ LTR. Upon reverse transcription, this motif will be copied into the 5′ LTR as well, and the internal proviral fragment should be removable by Cre. We combined LoxP insertion with deletion of the LTR promoter. Consequently, no transcriptionally active fragment is expected to remain after Cre-mediated excision. To substitute for the LTR promoter, our constructs contain an internal promoter that is positioned opposite the LTR. We used the minimal CMV promoter, preceded by an array of seven tet-operator sequences. This hybrid promoter (TORE) was modified from the original tetracycline-regulated promoter (16) by removing transcription factor-binding sites that were inadvertently introduced into the original construct (17). When tet-activator proteins are present, TORE becomes extremely powerful, and is stronger than the CMV immediate early promoter and the Moloney murine leukemia virus LTR (data not shown). A strong promoter oriented opposite to an LTR attenuates the function of the latter (8, 18). In the absence of a tet-activator in packaging cells, the interference of TORE with the LTR-dependent virus production is expected to be less than that from a constitutive promoter of comparable strength. On the other hand, inactivation of the LTRs and expression of a tet-activator in target cells would permit full use of the strength of the TORE.

Introns constitute a large portion of mammalian genes, and therefore, are the likely targets of retroviral integration. To facilitate the production of protein products by intron-targeted inserts, we introduced an unpaired adenovirus splice donor site downstream of TORE. Earlier, successful insertional mutagenesis with vectors containing splice donor sites has been reported by others (4, 19). A short ORF was positioned between TORE and the splice donor to supply the hybrid RNA with a translational start site. The proper translational frame depends on the properties of the target and cannot be predicted a priori. Therefore, we generated vectors with all three possible frames to be used in parallel. The fourth construct lacks the ORF and the splice donor site to prevent exon skipping upon integration upstream of the coding region of a gene. Finally, a neomycin phosphoribosyl transferase expression cassette, useful for selection in both bacterial and mammalian cells, was included.

Transfection of the constructs into Bosc23 packaging cells (20) produced virus stocks with ≈10⁴ colony-forming units per ml, as determined by serial dilution on HCT9 cells.

Reversible Promoter-Insertion Mutagenesis of NF-κB-Dependent Signaling. We used a selective system established in our laboratory. 293ZeoTK cells were derived from the HEK293 cell line by introducing a functional IL-1 receptor and NF-κB-dependent expression constructs for bleomycin-binding protein and thymidine kinase. Because their endogenous NF-κB activity is low, these cells are constitutively resistant to GCV and sensitive to bleomycin analogs (e.g., Zeocin). This resistance pattern can be reversed by treatment with cytokines, such as IL-1, or by mutations that lead to the activation of NF-κB (Fig. 7, which is published as supporting information on the PNAS web site). These cells were used previously to obtain mutants in NF-κB-dependent signaling (7, 14, 21). We introduced a tet-activator protein (16) and murine ecotropic receptor (22) into these cells to allow tet-OFF regulation and ecotropic infection, respectively. The resulting cell line was designated E8. E8 was passaged in the presence of GCV to eliminate spontaneous mutants with constitutive NF-κB activation.

In a test experiment, we used one of the splice donor-containing constructs. The corresponding virus was applied to E8 cells by using three consecutive rounds of infection. The infection efficiency was not determined because of the presence of a neomycin resistance gene in the target cells. Infected cells were grown in the presence of Zeocin. The yield of resistant clones was comparable to that seen with uninfected cells (10^-6 to 10^-5), indicating that the incidence of insertional and spontaneous mutants was comparable, and underscoring the need for an efficient procedure to separate the two classes of mutants. Eight independent mutant clones, distinguished by drug resistance profiles or patterns of insertions (from a Southern analysis, data not shown) were classified as reversible or nonreversible, and as cis or trans. Mutations affecting the DNA at or near the integration site of the resistance marker are called cis. They are distinguished by failure to affect the function of the second marker (thymidine kinase). Trans mutants, in which diffusible factors are altered, impact both marker cassettes and may be expected to affect components of the NF-κB complex itself or one of its numerous regulators. Induced trans mutants differ from spontaneous mutants by the dependence on the status of the inserted promoter.

To classify the mutants, we introduced either an expression construct for Cre or an empty vector control and tested the cells for resistance to Zeocin or GCV. Of the four clones that showed Cre-dependent reversion, two were cis and two were trans (e.g., Fig. 2C). We confirmed the induction of NF-κB in the reversible trans mutants by EMSA (Fig. 3A) and luciferase reporter assays (Fig. 8, which is published as supporting information on the PNAS web site). Also, the expression of the NF-κB transcriptional target IL-8 (Fig. 3B) was elevated reversibly. In the reversible trans mutant 3B22/35, Southern analysis revealed multiple integration sites (Fig. 4A). Comparison of the insert pattern in Zeocin-resistant and -sensitive cells revealed the association of one of the integration sites with the mutant phenotype. (Figs. 2C and 4A). This experiment gave the approximate distance between the insert and the next ApoI site in the host DNA (Fig. 4B). We used this information to identify the corresponding product of iPCR (23), which has been cloned and sequenced, revealing an insert in the first intron of the relA gene, which encodes the NF-κB subunit p65. The inserted promoter is colinear with the target gene. The fusion is expected to substitute 10 vector-encoded amino acids for the two N-terminal amino acids of p65 (Fig. 9A, which is published as supporting information on the PNAS web site). To confirm the results of iPCR, we amplified genomic DNA of clone 3B22/35 with a primer anchored to the third exon of relA and a primer annealing within the vector, and verified the insertion site by sequencing the PCR product (data not shown). Moreover, we confirmed the reversible overexpression of p65 by Western analysis (Fig. 4D), by RT-PCR with a vector-specific and a relA-specific primers (Figs. 4C and Fig. 9A), and also by Northern analysis with a relA-specific probe (Fig. 5A). Expression of relA was also reduced upon suppression of the regulated promoter by doxycycline (Fig. 5A).

Another reversible trans mutant, 3B3-11, contained a single insert (data not shown). The target of integration was identified as act1, a gene previously discovered in our laboratory in a cDNA library based screen for NF-κB activators (21). The promoter construct is colinearly integrated in the first intron. We have confirmed the overexpression of an act1 hybrid transcript by Northern and RT-PCR analyses (Fig. 5 A and B). Sequencing the RT-PCR product revealed that the hybrid mRNA encodes an ACT1 protein with 17 amino acids added to the N terminus of the shorter isoform (Fig. 9B).

Positive-Feedback Loops in NF-κB Activation. The discovery of relA as an integration target raised the question of how the overexpressed protein is activated in the mutant cells. Full activation of classical NF-κB involves phosphorylation of the p65 transactivation domain. Interestingly, in the relA mutant the increase in phosphorylated p65 is reversible by Cre, but appears to exceed the increase in the total protein (Fig. 4D), suggesting that the upstream components of this signaling cascade are activated, and that this phenomenon depends on p65 overexpression. We reported previously that enhanced secretion of cytokines can cause the activation of NF-κB in mutant- and tumor-derived cells (14, 24). Because several different cytokines are under NF-κB transcriptional control, a positive-feedback loop can be established. We used an NF-κB reporter cell line (13) to test whether medium conditioned by the mutant clones contains NF-κB-activating factors. Treatment of the reporter cells with medium conditioned by 3B22/35 cells caused a robust elevation of luciferase activity (Fig. 5C). Medium conditioned by 3B3-11 cells also activated NF-κB, but not as well as medium from 3B22/35 cells (Fig. 5C). Interestingly, this difference correlates with the magnitude of overexpression of NF-κB transcriptional targets in these cells (Fig. 3B) and the ability to activate a transfected reporter (Fig. 8). The ability to secrete NF-κB-activating factors was reduced sharply in both Cre-treated cell populations (Fig. 5C). We conclude that the secretion of NF-κB-activating factors by mutant cells occurs in an NF-κB-dependent manner.

Discussion

Forward genetics allows one to identify factors that are important for a biological phenomenon without prior knowledge (2). This task is complicated in diploid somatic cells, where a mutation cannot be mapped by genetic crosses and has to be either homozygous or dominant to become penetrant.

We now describe a modified insertional mutagenesis approach for forward genetics that combines important features of library based techniques with those of traditional insertional mutagenesis, while avoiding the limitations of either approach.

As in the other forms of insertional mutagenesis, integration sites can be identified readily by using iPCR or other methods. Unlike library screening, our approach relies on a nearly universal set of vectors that can be used in diploid cells of virtually any origin to target essentially any gene. The latter advantage sets our method apart from the use of even the most comprehensive libraries constructed from known cDNAs or the corresponding small interfering RNAs. Obviously, our delivery method could be improved further and could be tailored to specific experimental goals. In addition to the Moloney murine leukemia virus-based constructs described here, other methods based on transposons or lentiviruses, to achieve mutations in a cell-autonomous manner or in poorly dividing cells, could be beneficial.

Depending on the site, the hybrid transcript from an inserted promoter may be equivalent to full-length, truncated, or antisense (Fig. 6). The procedure is potentially capable of producing both gain- and loss-of-function events within the same experiment. Such a broad scope decreases the amount of necessary preliminary data and the need for complementary selection experiments aimed at the discovery of genes with opposing functions. Moreover, a functional hybrid transcript will normally confer a dominant mutant phenotype, making the method generally applicable to diploid cells and even to cells with a higher ploidy, making the phenotype depend on the function of the inserted promoter. Retention of an intact allele means that the mutant phenotype can be reversed upon promoter inactivation, even if the latter does not restore the function of the targeted allele.

The introduction of site-specific recombination for validation distinguishes our method from those previously reported (e.g., ref. 4), and is especially valuable when multiple inserts are analyzed within the same clone. High multiplicity of insertion is highly desirable because it decreases the number of cells needed for a given experiment. Moreover, because the yield of spontaneous mutants is a function of the total cell number, the relative incidence of induced mutants increases with more insertions. For the 3B22/35 clone, we would not have been able to identify the relevant insert solely on the basis of phenotypic reversion because all copies of the promoter are likely to be regulated similarly. On the other hand, preferential loss of one of the inserts in reverted cells validates the significance of a specific target.

A possibility not yet fully explored in our work is to use the genomic sequence adjacent to the integration site as a traceable tag for that insert. The frequencies of such fragments in an infected population would change upon exposure to various growth conditions, reflecting selective advantages or handicaps conferred by corresponding mutations. One could monitor this pattern, for example, as a collection of bands in a Southern experiment, as shown in Fig. 4A, or as a collection of iPCR products.

A test of our method in 293ZeoTK cells has been successful on several counts. First, in a small-scale experiment, we were able to identify three predicted classes: cis mutants, and trans mutants in NF-κB components (relA) and in NF-κB regulators (act1). Second, in the case of relA, we were able to identify which of the multiple inserts determines the phenotype without having to characterize each individual insertion site. Third, the selected mutants represent an ideal system to study the properties of the target genes because their phenotypes can be regulated. In fact, the relA insertional mutant has already provided important evidence for a positive-feedback loop involving secreted factors and NF-κB. We have previously reported that elevated secretion of certain growth factors can cause constitutive NF-κB activation in many cancer cell lines, and probably, also in tumors (13, 14). Others have reported that Ap1 causes elevated growth factor secretion and the ensuing activation of NF-κB which, in turn, cooperates with Ap1 (25). In that case, however, the cycle only partially depended on NF-κB because its inhibition failed to curtail Ap-1 activity. On the other hand, in the case of the 3B22/35 mutant, reversion of p65 overexpression sufficed to greatly reduce the secretion of NF-κB-activating factors. We propose that elevated levels of p65 decrease the efficiency of negative regulators, such as IκB, allowing for positive feedback. Because NF-κB activation is likely to help at least some cancers to survive, it is important to know whether the overexpression of p65 itself, followed by its constitutive activation by secreted downstream targets, can occur in cancer.