- Made available online as an Accepted Preprint 6 December 2010
- Accepted Preprint first posted online on 6 December 2010
Novel approaches to in vitro transgenesis
- Faculty of Life Sciences, Manchester Interdisciplinary Biocentre, University of Manchester, 131 Princess Street, Manchester M1 7DN, UK
1Developmental Biomedicine Research Group, University of Manchester, AV Hill Building, Manchester M13 9PT, UK
- (Correspondence should be addressed to A D Adamson; Email: antony.adamson{at}manchester.ac.uk)
Abstract
The study of gene expression is a major focus in biological research and is recognised to be critical for our understanding of physiological and pathophysiological processes. Methods to study gene expression range from in vitro biochemical assays through cultured cells and tissue biopsies to whole organisms. In the early stages of project development, considerations about which model system to use should be addressed and may influence future experimental procedures. The aim of this review is to briefly describe advantages and disadvantages of the existing techniques available to study eukaryote gene expression in vitro, including the mechanism of transgene integration (transient or stable), the different transgenesis systems available, including plasmids, viruses and targeted integration and knockin approaches, and paying particular attention to expression systems such as bacterial artificial chromosomes and episomal vectors that offer a number of advantages and are increasing in popularity. We also discuss novel approaches that combine some of the above techniques, generating increasingly complex but physiologically accurate expression systems.
Gene expression
Gene expression and function is regulated at a number of levels: 1) from an initial stimulus or signal (which can be either intracellular or extracellular, leading to activation of a signalling pathway); 2) alterations in the chromatin architecture of the DNA and the promoter sequences; 3) transcription and mRNA production (splicing, transport, stability and degradation); and 4) translation and the formation of the protein product (which may be subjected to post-translational modification). This review will focus on the ectopic expression systems that are currently available to utilise transgenesis and to study the regulation of gene transcription.
Endogenous versus ectopic gene expression
Specific RNA production by cells has long been held as a direct indicator of gene expression. Techniques such as northern blotting (Alwine et al. 1977) and RNAse protection assays (Sambrook et al. 1989) can be used to assay RNA levels within a population of cells, and nuclear run-on assays (Gariglio et al. 1981, Brown et al. 1984) have been used to measure transcription initiation. Later, reverse transcription (transcriptase)-PCR (RT-PCR), and more quantitative variants (real-time PCR and quantitative PCR) have been developed to analyse RNA expression. Each of these techniques relies on endogenous gene expression within the cell, and has their own advantages and disadvantages. However, commonly no suitable cell line exists that expresses the gene-of-interest. Furthermore, these assays measure gene expression by extracting RNA from a population of cells, thus averaging gene expression across an entire population. Single-cell analysis using sophisticated microscopy has revealed startling heterogeneity of gene transcription between cells even within clonal populations (Takasuka et al. 1998, Ashall et al. 2009, Eldar & Elowitz 2010). Analysis of gene expression at the single-cell level using microscopy requires the addition of an appropriate marker protein or dye to gene expression vectors in order to visualise these effects. This is most commonly assessed using reporter systems that can be ectopically expressed in the cell.
At the beginning of any project, critical decisions should be made to define the methodological approaches to be pursued. Considerations should include
-
The cell line(s) to be used.
-
The reporter gene(s) to be used (colourimetric enzymes, luminescent enzymes, fluorescent proteins, etc.).
-
Whether an increase in the copy number of a gene may lead to aberrant effects (e.g. gene overexpression, perturbation of feedback mechanisms, etc.).
-
What kind of delivery system would be most appropriate (plasmid, viral, episomal, stably integrated or transient expression).
These above considerations are not mutually exclusive and can often overlap. In this review, we shall identify the individual problems and advantages inherent to a range of experimental approaches used to analyse transgenesis and ectopic gene expression.
Choosing a suitable cell model system
The choice of cell line(s) should be determined early in any project, initially by analysing expression of the gene-of-interest in a number of relevant cell lines. For many genes, particularly endocrine genes that tend to be tissue- and cell-specific in their expression pattern, it is common to fail to identify a cell type suitable for gene expression analyses. For example, many endocrine genes are specifically restricted in expression to the pituitary, but their study is hampered by the absence of human pituitary-derived cell lines, and primary tissues rarely become available due to a lack of pituitary surgical intervention.
In the absence of human-derived cell lines for a specific investigation several alternative approaches to study gene regulation have been used. These include transient expression of specific transcription factors in HeLa cells (Bodner & Karin 1987), use of human cell lines isolated from different tissue origin that unusually expresses gene-of-interest (Gellersen et al. 1995) or the use of cells derived from the appropriate tissue, but from an alternative species, such as rodents (Ben-Jonathan et al. 2008). However, these approaches are not without potential criticism. It is difficult to account for cell-specific effects either lost or gained by using cells of different origins. Regulation of a specific gene may differ from species to species. For example, prolactin, which is expressed chiefly in pituitary lactotroph cells, has extensive differences in gene organisation and hormone function between humans and rodents (Gerlo et al. 2006, Ben-Jonathan et al. 2008, Bernichtein et al. 2010). In these circumstances, ectopic expression of human gene–promoter constructs in rat pituitary cells is a common approach for assessing the regulation of the human prolactin gene (e.g. Adamson et al. 2008, Semprini et al. 2009).
Additionally, cell lines have inconsistencies and limitations when compared with primary tissue. For example, many rodent pituitary cell lines lack functional dopamine receptors, and as a consequence are unable to respond to one of the most important in vivo pituitary regulatory signals. Although a lengthy and time-consuming process, it is possible to remedy this issue by generating stable cell lines expressing physiological levels of the two dopamine receptor variants resulting in a more physiologically relevant model system (An et al. 2003).
Often, compromises need to be made when selecting an appropriate cell line. Cell lines differ according to a number of characteristics including their growth and proliferation rates, whether they are adherent, easily transfected or transduced and whether aberrant gene expression or chromosomal aneuploidy occurs. Single-cell time-lapse microscopy experiments are far easier to analyse if cells remain relatively static. Thus, on occasions the cell line selected may not necessarily be the most physiologically relevant and the compromises that are made need to be acknowledged.
Reporter genes
Studies of mammalian gene expression have relied heavily on the use of reporter genes that encode a measurable exogenous protein product that can give a quantitative index of the level of gene activation gene when linked to a given promoter. The library of reporter genes available has significantly expanded over the years, and includes enzymes such as β-galactosidase, which generate a colourimetric product or chloramphenicol acetyl transferase, which catalyses the synthesis of radiolabelled (and latterly fluorescent) product, to enzymes such as luciferase that catalyses the production of light from the substrate luciferin to give a measure of gene activation (Frawley et al. 1994, White et al. 1995). Luciferase can be used to analyse the induction of gene expression across a population of cells by assaying the whole cell lysate, and also to dynamically analyse gene expression in a single cell through microscopy, measuring the photon output from a single cell (Harper et al. 2010).
Also available are fluorescent proteins such as green fluorescent protein, originally derived from bioluminescent marine species (Tsien 1998, Shaner et al. 2007). Fluorescent proteins have revolutionised not only gene expression studies but, through fusion to the coding sequence of the gene-of-interest, have enabled visualisation of the behaviour of the protein within the cell over time (Rafalska-Metcalf & Janicki 2007, Spiller et al. 2010).
Currently, a multitude of vectors that express differently coloured fluorescent proteins are available; most are altered version of the natural DNA sequence (e.g. cherry, tomato; Chudakov et al. 2010). Fluorescent proteins display different properties, including variable brightness, ability to form monomers or multimers, photo-switching, photo-activation, photo-conversion, and pH and temperature sensitivity (Giepmans et al. 2006, Chudakov et al. 2010). Further modifications such as addition of PEST or CL-1 sequences (Rogers et al. 1986) shorten intracellular half-lives of fluorescent proteins to allow gene expression to be monitored more accurately. Increasing our understanding of dynamic and stochastic gene expression requires the facility to monitor rapid changes in the transcriptional rate of the gene-of-interest (Raser & O'Shea 2004). Accurate quantitative data for reporter gene synthesis and degradation is essential in order to incorporate mathematical modelling and systems biology approaches into studies (Finkenstädt et al. 2008). Thus, reporter gene development is an ongoing process and is important to ensure that reporter molecules can be selected and tailored to meet the requirements of specific gene expression studies.
Driving gene expression
Many studies designed to investigate protein behaviour within a cell utilise expression vectors with gene transcription driven by constitutively active promoters with a broad range of activity in most cells, such as the viral cytomegalovirus, Rous sarcoma virus and Simian virus 40 (SV40) promoters. The constitutive transcription from these promoters is by way of a concatenation of numerous different response elements and is suitable wherever the constant production of protein within a cell is necessary. However, these promoters do not express the protein appropriately with regard to the onset of transcription, induction by hormones or drugs and regulation by other factors that may constitute feedback loops. Furthermore, when some genes are expressed at incorrect levels it can lead to aberrant activation or repression of other genes. The use of the relevant promoter of the gene-of-interest greatly enhances the suitability of data generated. Promoter-mediated regulation is highly specific to each individual gene and can often involve multiple transcription factor binding sites found in a multitude of different locations relative to transcription start site. Transcription factor binding sites tend to be primarily within the proximal region of the gene for most 5′ sequences, but enhancers and locus control regions can be many kilobases away, in both upstream and downstream and even intronic sequences (Carroll et al. 2005, West & Fraser 2005). An in-depth knowledge of promoter structure can greatly influence the choice of expression system; this will be discussed in further detail later in this review.
Gene delivery: mechanism of transgenesis
Before analysis of protein or gene expression can be analysed, the molecular ectopic expression systems must first be introduced to the cell. This can be achieved through a variety of biological, chemical or physical techniques including, for example, viral transduction, cationic lipid delivery or electroporation (Kim & Eberwine 2010) but is ultimately determined by the cellular model system(s) chosen. Many different expression systems are available, but all can be loosely categorised as either purified circular bacterial DNA structures (i.e. plasmids, episomal vectors and larger structures such as cosmids and bacterial artificial chromosomes (BACs); see later) or viral vectors.
Plasmids, small circular bacterial DNA structures, are perhaps the most commonly used ectopic expression system. They are easy to purify from bacterial cultures by alkaline lysis, and can be readily modified through well-characterised molecular cloning techniques. Plasmids can be introduced into cells using chemical transfection (including a wide variety of specialist compounds that differ markedly in efficiency, cell-type specificity and cost) and/or physical transfection through direct manipulation of the cells (e.g. electroporation and microinjection among other techniques). BACs are similar in many ways to other bacterial vectors in their propagation and purification and have several advantages when compared with plasmids, but modification and transfection are more challenging than using plasmids (see later).
Viral transduction approaches exploit the natural ability of viruses to deliver genetic material into infected cells. A number of viral-based vectors are commonly used to ectopically express reporter constructs, including adenovirus (AV), adeno-associated virus (AAV), herpes simplex virus (HSV) and retroviruses (principally lentiviruses (LV)). Each viral vector has a number of advantages and disadvantages and this has been reviewed extensively elsewhere (Verma & Weitzman 2005, Osten et al. 2007, Howarth et al. 2010). In this review, we present a brief summary of the general properties of viral transduction and summarise the characteristics and properties of commonly used viral vectors (Table 1).
Properties of viral vectors
Standard molecular cloning techniques are first used to generate a gene-of-interest-containing vector in vitro in suitable viral construct(s) containing genetic information critical for the formation of the virus. Next, after the infection of a host cell, facilitating replication of the viral genome within the host cell environment, and formation and propagation of new virions, the cells are lysed or the virus-containing supernatant is removed, centrifuged and added to the target cell line to infect and transduce the target cells. The viral titre or yield obtained from the host cell differs according to the virus used and some viruses, such as AAV, often require helper virus expression to increase the titre. Thus, preparation for performing transgenesis is lengthier and more labour intensive compared with using a simple bacterial vector construct.
Furthermore, viruses vary in their ability to infect different cell types, but usually target both dividing and non-dividing cells (Pfeifer 2004). Viruses often have limited cloning capacity (4–7 kb), with the exception of HSV, which can hold very large (>100 kb) DNA inserts (Cuchet et al. 2007). One key advantage of using viruses (with the exception of AVs; see Table 1) is the relative ease of generating cells with an integrated transgene without the need for selection markers and clonal selection strategies.
Gene delivery for gene transcription analysis: transient or stable expression
Transgenes can be either transiently expressed for a period of hours to days or stably integrated into the target cell genome through appropriate transformation techniques and/or the use of selection markers. Stable integration of transgenes can be a time consuming and laborious process, requiring weeks to months to generate suitable clonal cell populations depending on the transgenesis method used, whereas transient expression can yield data in a relatively short time frame (for a comparison of the different transgenesis methods and their transient and stable effects see Fig. 1).
Transient versus stable transfection. Multiple methods for achieving transgenesis are shown. Column 1 – plasmids. Multiple copies found transiently, and multiple integration sites upon stable transfection. Column 2 – lentivirus (LV), adeno-associated virus (AAV) and herpes simplex virus (HSV). Transduction is long term and not of a transient nature. LV integrates with several copies; AAV and HSV are both non-integrative. Column 3 – adenovirus. Virus usually lacks replication ability, thus only transient transduction, at multiple copies, is achieved. Column 4 – bacterial artificial chromosomes (BACs). Transient expression results in low copy number within the nucleus, stable expression is integrative at a low copy number also. Column 5 – episomal vectors. Transient expression of several copies, vector associates with chromosome and nuclear matrix. For stable expression, the vector continues to associate with the chromosomal regions and nuclear matrix but does not integrate. Column 6 – gene knockin or targeted integration. These are stable only approaches and involve targeting either the endogenous gene locus or a predetermined locus.
Many authors have reported discrepancies between these expression systems. For example, the MMTV promoter, a sequence extensively used in transcriptional studies, varies greatly in activation by different transcription factors depending on transient or stable transfection (Archer et al. 1992, 1994). This was shown to be due to the incorrect nucleosome positioning of the DNA transiently transfected into cells (Hebbar & Archer 2008). It has been demonstrated that transiently transfected DNA often fails to efficiently form the appropriate chromatin structure and, in some cases, regulation of these transgenes may not accurately represent the in vivo behaviour of the promoter-of-interest. Some strategies exist to circumvent this issue by use of cell lines expressing SV40 T antigen and integrating the SV40 origin of replication into the vector. Using this approach, plasmid DNA is assembled into nucleosomes that mimic the endogenous gene, and the resulting mini-chromosomes are more accurately replicated and transcribed by the cellular machinery (Xu & Cook 2008). However, there are likely to be multiple copies of the transgene, which may not faithfully reproduce expression profiles.
Additionally, transient transfection efficiency also can differ extensively between experiments. A common practice is to co-transfect control reporter plasmids, to help determine the level of transfection between samples, although some authors report that regulation of these vectors themselves may lead to systematic error in the experimental data (Huszar et al. 2001). Furthermore, Ishikawa et al. (2004) have reported that co-transfected plasmids have the potential to form concatamers, through non-homologous end-joining ligation, leading to further problems in interpretation of the results.
Despite the problems associated with transient transfection, the technique remains a quick and simple method for analysis of gene transcription, particularly when screening a number of promoter–reporter constructs (such as in promoter deletion and mutagenesis assays). While stable cell lines are more likely to reflect the true nature of molecular interactions in gene regulation, especially with regard to chromatin and transcription dynamics, the generation of stable cell lines is time consuming and laborious and clones need to be well characterised. Furthermore, clones may suffer from ‘site of integration effects’ and copy number abnormalities (see below), and it may be advisable to experiment with a number of different clones, or a mixed population of clones, to ensure the activation of the target promoter is consistent throughout, but overall stable clones do provide a more consistent and well-characterised cell model system.
Increasing gene copy number
The introduction of reporter transgene(s) usually results (with the exception of gene knockins; see later) in altered gene copy number within a cell, the effects of which are understudied and, therefore, poorly understood. The most thorough analysis of how increasing the gene copy number affects expression in vitro has come from studies in yeast (Guan et al. 2007, Presser et al. 2008) and Drosophila (Sabl & Henikoff 1996). With regard to mammalian gene expression, transcriptomic data suggest that the number of transcripts from a gene roughly increases in proportion to gene copy number (Zhang & Oliver 2007). It is therefore possible that ectopic transgenesis approaches that result in increased gene copy number within the cell may perturb the biological system. It is relevant to mention that some studies adopting a mathematical approach have suggested that increased gene dose can be compensated for through proportional effects on feedback loops (Mileyko & Weitz 2010). Based on the above observations it is likely that the effect of copy number is gene and system specific. Thus, the effect of increasing transgene copy number remains an unknown quantity in most systems and needs to be acknowledged as different ectopic expression systems vary in the number of transgene copies they introduce to the cell. It should also be noted that due to epigenetic silencing of non-mammalian DNA (see ‘Integration effects’ and Fig. 2B) that the effective copy number of a clone may reduce over time with increasing generation number (Jenke et al. 2004, Mutskov & Felsenfeld 2004; Table 2).
Site of integration effects. Black lines/boxes/arrows represent endogenous DNA, red lines/boxes/arrows represent transgene DNA. (A) Integration of transgene into different chromatin states. Upper panel shows inhibited transcription of transgene upon integration into dense heterochromatin state (‘Me’ indicates hypermethylation), lower panel shows transgene able to be transcribed when integrated into open, euchromatin state. (B) Insertional mutagenesis; disruption of an endogenous gene. Integration of transgene disrupts a random endogenous gene resulting in loss or altered expression. (C) Integration within the range of endogenous transcriptional control elements, left panel shows overexpression of transgene due to the effect of endogenous enhancer, right panel shows inhibited transgene expression due to the effect of endogenous insulator. (D) Transgene silencing over multiple generations. Initial integration results in transgene expression but passaging results in the spread of heterochromatin, particularly if the exogenous DNA contains non-mammalian DNA.
Comparison of various ectopic expression systems
Integration effects
The site(s) within the genome at which an integrated reporter gene resides can have a profound influence on gene expression, either of the transgene itself or surrounding genomic regions. Integration of a transgene into dense, inactive heterochromatin regions can result in unstable position effect variation, resulting in little or no transgene expression (Fig. 2A; Dillon & Festenstein 2002). It is also possible that gradual silencing of the transgene can occur over time as heterochromatin can spread into the majority of integrated genes, especially when the exogenous DNA contains bacterial DNA sequences (Fig. 2B; Jenke et al. 2004, Mutskov & Felsenfeld 2004). Integration of exogenous DNA may also disrupt important endogenous sequences including genes and regulatory elements, leading to unanticipated changes in the cell phenotype (insertional mutagenesis; Fig. 2C).
Integration within the ‘range’ of endogenous enhancers/promoters may lead to overexpression of the transgene (Fig. 2D). Given that enhancers and locus control regions can influence genes found several kilobases away this ‘range’ is potentially quite large. The site of integration can be critical for correct control of gene expression and there are a number of different methods to limit these effects, including the addition of insulator elements to flank the transgene and its promoter (Fig. 3A, upper panel; Bell et al. 2001). Insulator elements are natural genetic sequences that establish independent domains of transcriptional activity within the eukaryotic genome and have distinct roles in establishing gene expression; acting as barrier elements to prevent the spread of heterochromatin and blocking enhancer–promoter interactions (Fig. 3A, lower panel; Kuhn & Geyer 2003, Recillas-Targa et al. 2004). Thus, while insulators would not protect against insertional mutagenesis that may lead to aberrant gene expression of non-transgenes (Fig. 2C), flanking transgene with insulators confers position independence, allowing self-regulation, appropriate chromatin organisation and prevents transgene silencing by the spread of heterochromatin. Insulators have proved highly effective both in vitro (Anastassiadis et al. 2002, Qu et al. 2004) and also when used in vivo in the development of gene therapy approaches (Gallagher et al. 2009).
Methods to minimise site of integration effects. Black lines/boxes/arrows represent endogenous DNA, red lines/boxes/arrows represent transgene DNA. (A) Flanking transgene DNA with insulator sequences. If the transgene integrates within the range of endogenous transcriptional regulatory elements, insulators can block their effects (upper panel). Insulators can also halt the spread of heterochromatic and gene silencing (lower panel). (B) Targeted integration. Following determination of recombinase target integration site, the gene-of-interest can be specifically targeted to a favourable locus. (C) Knockin. Owing to the targeting of the endogenous gene locus no site of the integration effects will be observed, either when a reporter is knocked in to replace the endogenous gene (upper panel), or when a fusion protein knockin is created (lower panel). (D) Use of large constructs. The extended flanking sequences found in BACs leave the transgene itself out of range of enhancers and insulators. (E) Use of episomal vectors. Episomal vectors associate with the nuclear matrix close to the chromatin through the scaffold/matrix attachment region (S/MAR) but do not integrate.
Targeted integration
Integration of transgenes into pre-determined endogenous genomic loci, and subsequent selection by an appropriate marker, known as recombinase-mediated cassette exchange (RMCE; Schlake & Bode 1994), is a method for generating stable cell lines and limiting site of integration effects (Fig. 3B). In this approach, a pool of stable cell lines is created that have recombinase target sites (including Flp recombination target (FRT), locus of X-over P1 (LoxP), and more recently phiC31 integrase sites; Oumard et al. 2006), integrated into the genome along with a selection marker. Screening of these founder clones can then identify the genomic position of these recombinase sites, and suitable clones can be selected and expanded. Criteria for clones include integration of a single copy, in a genomic locus of high transcriptional potential and a lack of insertional mutagenesis (Qiao et al. 2009). These clones can then be subjected to transfection with an expression vector carrying the gene-of-interest, and a complementary recombinase site. Transient expression of the relevant recombinase facilitates integration of the gene-of-interest into the pre-determined genomic loci (Thomas et al. 2004). This process can be engineered so as not to leave a selection marker or extensive prokaryotic DNA, reducing the likelihood of epigenetic silencing (Oumard et al. 2006).
A drawback to this approach is the random integration of the initial recombinase target site and clone screening, which may be time consuming and subject to chance. Once a suitable line is identified, however, it allows great flexibility and rapid and efficient generation of copy number-controlled stable cell lines in the future, especially useful if a number of constructs are to be screened (e.g. mutagenesis and promoter length studies). There are many easy-to-use commercially available kits for this process, as well as pre-generated recombinase target site cell lines, negating the first, lengthy step.
RMCE has numerous forms, each with their own advantages and disadvantages and varying levels of complexity (e.g. FRT site variants allowing for RMCE multiplexing; Bode et al. 1992, Turan et al. 2010) and has been expertly reviewed elsewhere (Oumard et al. 2006, Wirth et al. 2007). These methods provide high flexibility in experimental approach, but may require extensive molecular engineering contributing to generation time of the cellular model system, and concessions on the background cell line may have to be made.
Gene ‘knockin’
A gene ‘knockin’ can be specifically targeted to the endogenous gene locus to homologously recombine in the reporter construct of choice, either as direct readout of promoter activity (by replacing the gene itself with a reporter gene; e.g. Ishikawa et al. 2006) or combined with the endogenous gene as a fusion reporter (e.g. De Lorenzi et al. 2009). The advantage of this approach is the targeting of endogenous genes, resulting in no variation in gene copy number or site of integration effects, as well as the expression of the full gene (introns, exons, untranslated region), which contribute to gene regulation. This approach is commonly used in embryonic stem (ES) cells for the generation of knockin/reporter animals and in Saccharomyces cerevisiae in which the high homologous recombination rate in yeast has been exploited to generate epitope tags on a number of yeast genes and proteins (Gavin et al. 2006). Gene knockin approaches have also been applied to somatic human cell lines for in vitro gene analysis (Hendrickson 2008, Kim et al. 2008). However, homologous recombination in mammalian cells is a highly inefficient process (Yáñez & Porter 1999) and, in particular, human somatic cell lines have been shown to display an intrinsically low rate of homologous recombination (Sedivy et al. 1999, Hendrickson 2008). Lengthy homology arms (at least 2–3 kb) are necessary to increase recombination efficiency resulting in complex molecular construction.
Developments increasing the efficacy of this approach in somatic cells have proved invaluable. It has been shown that the delivery method of the recombination cassette DNA is critical. Hirata & Russell (2000) used a recombinant AAV approach that greatly increased recombination efficiency when compared with transfected plasmids (Topaloglu et al. 2005, Rago et al. 2007). However, again the cell model system was essential, as there was a huge variation in the efficiency of transfection and recombination between the cell types (Rago et al. 2007).
Furthermore, in addition to modifying the endogenous locus, the exogenous DNA can be randomly integrated at multiple, non-targeted sites, leading to both site of integration effects and an increase in transgene copy number. Sophisticated molecular approaches that increase targeting efficiency have been developed, including promoter traps where the transgene recombination cassette contains not only the desired modification and a selection marker flanked with loxP sites (or other suitable recombinase target) but also a splice acceptor immediately upstream of an internal ribosome entry site antibiotic-resistance gene (Topaloglu et al. 2005, Rago et al. 2007). The splice acceptor initiates transcription of the selection marker driven by the endogenous promoter of the gene, thus selection marker expression only occurs when correct DNA integration is achieved. The marker can then be removed through transient expression of the necessary recombinase (the proximity of a constitutive promoter-driven marker gene expression could conceivably have indirect effects on the regulation of the endogenous gene; Hasegawa & Nakatsuji 2002). However, it is important to note that the endogenous target gene must be basally expressed at a sufficient level in order to drive selection marker expression, potentially limiting the usefulness of this approach for some genes and cell types.
Creating knockin somatic cell lines is a lengthy procedure, requiring generation of complex targeting cassettes, followed by delivery of the cassette via a suitable method to a limited number of cell lines at low efficiency. Following this, stable clones need to be selected and carefully characterised. However, the lack of copy number alterations and the site of integration effects (Figs 1 and 3C) result in an excellent model system for in vitro molecular tagging gene expression analysis (Kim et al. 2008).
Large constructs
Another approach to minimise site of integration effects is to use large constructs such as yeast, bacterial or P1 artificial chromosomes (YACs, BACs and PACs respectively). YACs were originally developed for cloning large genomic fragments into yeast cells and BACs and PACs were developed for cloning large genomic fragments into Escherichia coli. YACs can contain huge, megabase-sized DNA inserts (compared to 100–350 kb inserts of BACs and PACs) but YAC DNA is notoriously difficult to purify. YACs exist in multiple copies in bacteria that cause chimaerism, and random recombination events occur frequently (Copeland et al. 2001). YAC manipulation by recombination is an even lengthier process than that with BACs/PACs; thus, YACs are not as popular a model system as their bacterial counterparts for transgene vector construction.
BACs can be considered large plasmids, propagated in bacteria and consisting of a small amount of bacterial DNA derived from the single copy F-plasmid (Shizuya et al. 1992) containing prokaryotic genes essential for replication, partition and selection, and a large genomic fragment of mammalian DNA (100–350 kb). These vectors were a useful tool in the genome-sequencing projects (Lander et al. 2001, Venter et al. 2001) and as a result most regions of the human genome (as well as genomes of several other species) are available as BACs. With the development of homologous recombination systems in E. coli (Copeland et al. 2001), BACs can be genetically engineered to express reporter genes, or to place a transgene under the control of inducible or conditional promoters making them an increasingly popular method of gene expression in the mammalian cells. The sheer size of BAC DNA vectors helps to minimise the site of integration effects (Fig. 3D).
In addition to minimising the integration effects, BACs have a number of other advantages. Often, proximal upstream promoter elements are used to drive gene expression and small constructs such as plasmids and viruses may have up to 5 kb of DNA immediately flanking the gene-of-interest to drive reporter gene expression. However, transcriptional regulatory mechanisms are highly gene specific and proximal-flanking DNA is often insufficient to drive transgene expression (e.g. human GH; Palmiter & Brinster 1986), or at least does not reflect the full promoter behaviour for gene activation (Semprini et al. 2009). In such cases, the use of larger constructs has shown the necessity of surrounding genomic sequences, containing enhancers and locus control regions, resulting in a more accurate representation of in vivo gene expression (e.g. Townes et al. 1985, Jones et al. 1995). Using a larger construct that contains a sizeable fragment of the human genome surrounding the gene-of-interest will therefore increase the probability that any such regulatory elements are included. Moreover, bioinformatic programmes are increasing in sophistication and regulatory regions such as enhancers and locus control regions can be predicted from the genomic sequence. This process can either be based on the identification of predicted or probable binding sites for specific transcription factors, or alternatively through the analysis of conserved non-coding DNA sequences (Sakabe & Nobrega 2010). Combining bioinformatic approaches with the selection of BAC clones allows the inclusion or exclusion of potential regulatory sequences.
In recent years, there has been increasing interest of the role of RNA in gene regulation, including RNA splicing and gene regulation by microRNAs (Jackson & Standart 2007). As BACs are essentially a fragment of the human genome they will contain the full gene structure, including untranslated regions, exons and introns, alternative promoters and splice sites and microRNA coding sequences. This should result in full mRNA processing and splicing when transcribed, and will also include any microRNA targets sites within the mRNA sequence and produce the full complement of protein isoforms once translated. Thus, genes expressed from BACs mirror endogenous gene expression far more accurately.
Despite their size, it is still possible to transiently express BACs in mammalian cell lines, though the efficiency of transfection tends to be lower than for smaller plasmids (Magin-Lachmann et al. 2004) but they are introduced into cells at a lower copy number (Sparwasser & Eberl 2007).
BAC recombineering technology includes different selection strategies that can minimise any disruption to genomic sequences within the BAC. Some methods can cause minimal alterations to a BAC sequence and leave a post-recombination scar, usually comprising recombinase target site such as FRT or loxP. Other methods use selection/counter selection, which results in a seamless alteration of the BAC, allowing for accurate manipulation without altering neighbouring sequences (e.g. the GalK system described by Warming et al. (2005)). This is important in the generation of fusion protein, expressing BACs where the reading frame of the insertion is critical. A fusion protein might not behave naturally; therefore, transient transfection of the BAC construct should be performed initially to confirm expression of the gene-of-interest and the correct behaviour of the protein before undertaking the time-consuming step of stable clone generation. It is also advisable to perform initial studies with plasmid vectors that express the same (or very similar) fusion proteins before embarking on BAC construction.
There are a number of disadvantages associated with using BACs: i) a construct consisting of a large genomic fragment is likely to contain non-related genes that may lead to indirect, non-specific gene expression and potentially undesirable phenotypes; ii) the generation and screening of recombinant BAC constructs can take considerably more time than that of plasmids or other gene expression vectors; iii) large DNA constructs require special handling as they are more susceptible to degradation and shearing; and iv) recombination systems may suffer random recombination events, for instance the problems associated with cryptic LoxP sites resulting in random Cre-mediated recombination (Semprini et al. 2007). Moreover, BACs that contain repeating homologous sequences are prone to intramolecular rearrangements, not only reducing the efficiency of recombination but also, in some selection/counter-selection approaches there are a high number of false-positive clones, increasing the time spent screening (Narayanan 2008).
In summary, BACs have numerous advantages over conventional plasmids in that they insulate the gene from the site of integration effects, allow effective chromatinisation of the promoter DNA and include distal regulatory regions within the construct. Expression driven by the genomic region in which the gene is located results in accurate transcription regulation and promoter activity and promoter feedback. However, given their size, the technical difficulties in handling them properly and the potential for non-related gene expression, careful consideration of the investment in both time and resources for using BACs as a gene expression model system must be addressed.
Episomal gene expression systems
Integration effects can be avoided altogether through use of extra-chromosomal/episomal vectors designed to replicate independently of the genome (Fig. 3E). These systems fall into three categories: mammalian/human artificial chromosomes (MACs/HACs), viral episomal expression systems and vector-based episomal expression systems (Lufino et al. 2008).
MACs/HACs are chromosomal-derived vector systems, which replicate autonomously, segregate into daughter cells and are maintained at a low copy number (Conese et al. 2004, 2007). However, construction of these vectors is very difficult due to their large size and the time-consuming recombination techniques required, thus limiting the use of MACs and HACs as ectopic expression systems.
Many episomal systems are derived from viral sequences, a variety of which have been manipulated. The most successful among these has been the HSV-derived constructs, which are well retained through generations and have a relatively large insert capacity (Lufino et al. 2008).
The original EPI vector (pEPI-1) was constructed by Piechaczek et al. (1999) and contains chromosomal scaffold/matrix attachment region isolated from the human β-interferon gene cluster (Bode et al. 1992). This region interacts with major nuclear matrix proteins (Jenke et al. 2002) and this is thought to enable co-replication with chromosomal DNA. The vectors are mitotically stable and exist at a copy number of around 5 to 15 copies per cell (depending on the stage of the cell cycle; Stehle et al. 2003), which is probably relatively low when compared with transiently transfected plasmids but compared to stably integrated plasmid vectors will not have aberrant site of integration effects on the host. Stable transfection with EPI vectors is often highly efficient, compared to the integration of conventional plasmid vectors. EPI vectors have also been shown to remain in cells in the absence of selection for several hundred generations (Piechaczek et al. 1999, Baiker et al. 2000) probably because there is no chance of heterochromatin spread.
Combining approaches for transgenesis
In recent years, attempts have been made to combine different techniques to minimise and negate some of the disadvantages, while exploiting the various advantages.
Targeted integration utilising AAV
As discussed, one of the major drawbacks of targeted integration is the random integration of recombinase target sites when generating ‘founder’ cell lines prior to integration of the gene-of-interest. DeKelver et al. (2010) have utilised the natural ability of AAV to target and integrate in a specific genomic locus on chromosome 19 (commonly referred to as AAVS1 locus; McCarty et al. 2004). Although integration at this site disrupts the PPP1R12C gene, this does not alter cellular phenotype (Smith et al. 2008, Hockemeyer et al. 2009) nor has any pathophysiological effect from AAV infection been observed. DeKelver et al. (2010) thus report this particularly well-characterised genomic region to be a transgene ‘safe harbour’, as no phenotypic effect has been observed by its disruption, yet the region is transcriptionally active. The authors exploited this natural AAV integration site to generate somatic cell lines with easily targetable recombinase sites at this locus. Founder line integration site is therefore consistent across cell types, thus reducing the time and effort necessary for cell line screening and characterisation prior to targeted integration of the gene-of-interest.
Using BACs for targeted ‘knockin’
Creating knockin and knockout cell lines is a lengthy and challenging process. One of the key components of the development is the generation of the homologous recombination cassette, usually several kilobases in length. Recently, Song et al. (2010) exploited the coverage of the human genome by BACs to generate a recombination cassette with extended homology arms for the creation of p53 knockout human ES cells. This BAC-based targeting approach had several advantages, such as the introduction of large DNA sequences, leading to a high efficiency of homologous recombination in various genetic backgrounds. This technique avoided the more complex and technically difficult cloning strategies by using BAC recombineering technology to generate the recombination cassette (Song et al. 2010). One reported drawback of this BAC-based targeting approach was the difficulty in confirming that the homologous recombination event had occurred. It should also be noted that ES cells are more amenable to such an approach compared to somatic cell lines. Thus, generating a recombinant BAC in vitro can improve targeted integration and recombination in in vitro cell cultures.
Episomal BAC expression
When producing cell lines for study, the uncontrolled integration of BACs into genomic DNA could lead to (albeit minimised) the site of integration effects, the BAC might be truncated unpredictably and the low transfection efficiency of BACs makes generating stable cell lines challenging.
Utilising BAC technology in combination with episomal sequences might be an effective way to stably introduce BAC constructs into the mammalian cells. Episomal constructs replicate autonomously within the cell and do not integrate in the genome, in addition to being relatively simple to use for generating stable cell lines. Lufino et al. (2007) were able to fuse the pEPI vector (Stehle et al. 2003) with both HSV viral sequences and a BAC to express the low-density lipoprotein (LDL) receptor in a cell line that was deficient in this gene. Subsequent viral infection of this construct resulted in LDL receptor-expressing cells, with the construct maintained at a low copy number (∼2) in most cells for many generations. Thus, the combination of these two strategies led to exploitation of the major advantages of both systems and reduced the impact of the respective disadvantages for both techniques. Furthermore, the presence of HSV sequences within the EPI-BAC allows viral transduction of the construct, circumventing the problematic issues with BAC transfection. This also negates the need for special BAC DNA handling in the laboratory, and results in increased efficiency for delivery of DNA to target cells.
Conclusions
The study of gene expression is an ever expanding field of research, with a variety of technologies and approaches available. Important decisions are required when selecting a model system for analysis in any given project. All approaches discussed have their advantages and limitations, but methods are constantly evolving and adapting, expanding the options available to researchers.
Declaration of interest
The authors declare that there is no conflict of interest that could be perceived as prejudicing the impartiality of the research reported.
Funding
The work by the authors was funded by the Wellcome trust (grant 67252) and the BBSRC (grant BB/F00561X/1).
Acknowledgements
We thank Prof. M R H White for the support and valuable discussions. We thank Dr D G Spiller for critical reading of the manuscript and comments.
- Received in final form 23 November 2010
- Accepted 1 December 2010
- © 2011 Society for Endocrinology
