We have described an efficient technique which allows the selection of the stable integration of an exogenous transgene into the host genome. The importance of this study is twofold:
we have demonstrated that the versatility of the Tol2 system can be extended to normal chicken erythroid progenitor cells and
we have demonstrated the feasibility of combining this chromosomal integration technique with a MACS (MAgnetic Cell Sorting)-based enrichment method, allowing the instantaneous sorting of cells carrying the transgene.
The value of the Tol2 transposon-derived transgenesis strategy has been confirmed by reports of its ability to undergo efficient transposition in a wide variety of vertebrate species [7–12]. Among the DNA transposable elements, the Sleeping Beauty transposon, a member of the mariner family , has been previously reported to be capable of undergoing efficient transposition in chicken . However, a Sleeping Beauty-based transgenesis system was found to be inefficient in primary avian cells, despite its efficacy when tested on immortalized avian cells . The present report demonstrates that Tol2, contrary to Sleeping Beauty, provides an efficient transgenesis system in primary avian erythroid progenitors.
Regarding the optimal molecular ratio between the expression and the helper plasmid, our data suggest that transposition is more efficient when cells are cotransfected with more of the Tol2 expression plasmid than the transposase one. This observation seems to indicate that there is an "overproduction inhibition", which leads to a decrease in transpositional activity in the presence of an increased amount of transposase . Nevertheless, it has recently been reported that the Tol2 transposon-based system does not exhibit overexpression inhibition within the tested transposase concentration range . Even if, at first sight, there seems to be a discrepancy between this report and our results, there is a subtle experimental distinction which may explain this difference. In reference , the total amount of DNA was variable. This is in contrast with our own study in which we decided to fix the total DNA amount. The increase in the level of the transposase plasmid was then accompanied by a decrease in the amount of expression plasmid. Moreover, in our cellular model the efficiency of the Tol2/MACS system is proportional to the quantity of Tol2 expression plasmid present, as well as the number of cells to be transfected (see Additional File 3). Therefore, the reason for the decreased transpositional efficacy observed with the 1/5 ratio is likely to be the decrease in the amount of Tol2 expression plasmid rather than an "overexpression inhibition".
In our experimental model, we were limited both by a maximal total of DNA used for the transfection and by the number of cells to be transfected, since the cell viability after nucleofection is inversely proportional to these parameters (see Additional File 3). These features are inherent to our model (i.e. primary erythroid progenitors and the nucleofection technique) so, when transferring the Tol2/MACS method to other cellular models and/or other transfection techniques, this should be set and optimized in order to take full advantage of this method.
We developed an approach based on the splinkerette technique  that allowed us not only to prove that there is an integration into the cellular genome, but also to actually characterize the insertion sites. Our results show that there is one single insertion point per T2EC clone analyzed, suggesting that the transposition event is a quite discrete event which probably leads to only one (or at most very few) insertion site. The single insertion point is localized on different chromosomes in the two analyzed clones. Whether this insertion site localization is random or whether there are some preferential genomic regions cannot currently be determined.
In order to obtain a population where a significant proportion (if not all) of the cells express the transgene, it is necessary to select or sort the transfected cells. When working on established cell lines, a common method consists of establishing transgenic clones or lines, which can be time-consuming. Although working ex-vivo with primary cells allows for more relevant studies (because they are normal and non-genetically altered cells), the lifespan of such cells is, by definition, limited. Hence we have chosen the MACS technique as it allows instantaneous enrichment of the population of cells carrying the transgene, unlike selection-based techniques which are longer and often introduce a bias due to their high selectivity. For example, it has been reported that selecting antibiotic-based agents can alter cellular metabolism genes . Our data shows that the present MACS-based method leads, after three successive selections, to a population in which about 70% of the cells stably express the transgene. Previous work in the laboratory  showed that transitory transfection of SCA2, a gene involved in the self-renewal of T2EC, has significant biological effects for up to three days after transfection. As these progenitors have a rapid cycle (they divide every 18 hours ), the transgene expression decreases in the whole population. In fact, cells carrying the transgene are diluted in a population that has doubled its size four times in three days. Despite this, three days after SCA2 transfection a significant effect can still be observed, which demonstrates that this effect can be detected even if there is only a small fraction of cells expressing it. Therefore, having 70% of cells expressing the transgene in a stable manner within the population is sufficient for carrying out studies on the genetic functions in these cells.
The efficacy of the MACS system relies upon the expression of the cell surface marker ΔhCD4, the translation (and, hence, expression) of which depends on the efficacy of the IRES. It has been shown that IRES efficiency is lower than that of the "classic" ribosome entry site (at the mRNA 5' extremity, which drives the cap-dependent translation) . Characterization of the expression of this marker, compared with the transgene expression reported by the eGFP, confirmed this assumption. Therefore, when using the present method, one should take into account that about half of the positive cells will actually appear as false-negatives (i.e., expressing the transgene but not the marker). Moreover, the expression level of ΔhCD4 is an important parameter, since the efficacy of positive cell retention on the magnetic column depends on the amount of microbead-conjugated antibodies at the cell surface, which is proportional to the expression level of this surface marker. In our experiments, the first MACS selection (leading to almost 100% positive cells) was carried out between 14 and 18 hours after transfection, since we observed that a MACS cell sorting performed more than 24 hours later was remarkably less efficient (data not shown). Timing is, therefore, an important parameter since we observed that the earlier the selection is done, the higher the expression level is and, consequently, the higher the efficiency. According to these observations, it is tempting to argue that the time point we use is so rapid that we are likely not to select stable integration events. However, our data shows that even if the first two MACS rounds of separation are equally efficient on cells expressing the transgene stably or transiently (with or without transposase), the third one is only efficient when cells are cotransfected with transposase and, hence, stably express the transgene. This suggests that these first two MACS selections are needed to increase both the proportion of cells expressing the transgene and the expression level of the ΔhCD4 surface marker in order to increase the efficacy of the third, and last, MACS separation, which is in fact the most decisive.
Furthermore, the fact that cells transfected with the empty plasmid also express the ΔhCD4 marker shows that the absence of a gene cloned upstream of the IRES does not impair its expression. This is an important feature since, in functional studies based on transgene expression, the negative control is often the condition transfected with the same empty plasmid. Therefore, the present strategy allows the sorting of cells transfected with the empty plasmid in the same way as cells transfected with the transgene-containing plasmid and, hence, provides a pertinent minus control. We also observed that, even though it is marginal, there is a small fraction of cells that would appear as false-positives and could, therefore, be selected. This may explain, at least partially, the fact that after each MACS step there is always a fraction of negative cells remaining (not expressing the transgene). However, this could also be explained by the extinction of the transgene expression. As both the IRES efficacy and the specificity of labeling with the microbeads-conjugated antibodies depend on the cell type, this should be tested in a preliminary experiment when transferring the current method to another experimental model.
According to the ΔhCD4 expression data, it is tempting to suppose that FACS (Fluorescence-Activated Cell Sorting () is much more efficient than MACS, but even if there is a difference in performance it is not particularly striking. Supposing the same method was adapted to FACS (i.e. with a fluorescent reporter gene downstream of the IRES instead of the ΔhCD4), there would also be a fraction of false-negative cells since the level of reporter gene expression depends on the efficiency of the IRES. Therefore, the proportion of true-positive cells (i.e., expressing the transgene and the reporter fluorescent gene) would not be very different from the one observed with the ΔhCD4, which is about 20% in T2EC. In practice, for MACS, we determined in our experimental model that up to 10% of the cells are actually sorted (see Additional File 4) This observation suggests that among the 20% of cells expressing the ΔhCD4, the magnetic column will actually retain only half. FACS is based on optical properties detected by an instrument, whereas MACS depends on several physical interactions (ΔhCD4 marker/antibody/magnetic column), so it is possible that FACS would have a higher sensitivity and this could increase the percentage of sorted cells up to 20%.
Nevertheless, if this constitutes an obvious advantage of FACS over MACS, there are also many advantages of MACS over FACS. First of all, FACS requires the expression of a fluorescent reporter gene, which is generally present in the cytoplasm. The expression of the MACS-required ΔhCD4, which is only at the cell surface without any cytoplasmic region present (as it has been deleted), is therefore more confined and more silent as regards cell function. Also, the antibody-conjugated microbeads are ultrasmall (approx. 50 nm ∅), biodegradable, and non-toxic to cells. Thus, transfected cells are magnetically selected without affecting cell function or viability. Furthermore, from a purely practical point of view, the MACS technique and equipment are more accessible than those used for FACS. In fact, FACS requires the appropriate sorter, an instrument that may not be available in the laboratory. On the other hand, the MACS equipment is very simple: the magnetic sorter can be placed on the lab bench as it is as easy to handle as a normal test tube rack. Moreover, whereas FACS selection can be time-consuming, since it sorts cell by cell, with a MACS experiment it is possible to instantly sort up to 107 positive cells, within 2 × 108 cells per column, in about one hour (including the labeling and wash steps), with the column version used in our experiment. For cells which have a reduced viability outside of the incubator, this could make a huge difference. In addition, it is possible to work with several columns at the same time, or to work with high capacity columns (up to 108 positive-cells within 2 × 109 cells per column).
However, it is important to note that our system is compatible with a Tol2/FACS variant. Instead of using the microbead-conjugated antibody and a magnetic column, it is possible to use a fluorochrome-conjugated hCD4 antibody (equivalent to the one used to characterize the expression of the ΔhCD4 in this study) to label the cells carrying the transgene and then sort these cells according to their fluorescence.
The MACS technique provides an efficient, rapid, gentle and easy way to sort cells, but methodological improvements could still be made. We believe that since the efficiency of the IRES-based expression might be questionable (see above), the priority is to develop an alternative Tol2/MACS system without the drawbacks of the IRES. We are, therefore, currently designing a Tol2 construct with two genes (the one of interest and the ΔhCD4-coding sequence) driven by two different promoters, leading to higher expression on the ΔhCD4 and, hence, a higher MACS efficiency. However, it is known that constitutive and highly active promoters can mutually interfere with each others functioning . This is why it seems relevant to add, between the transgene and the sorting gene, an insulator  to block any interference between the two promoters. The functional analysis of such a construct is, however, beyond the objectives of the current study.
It is, therefore, expected that further improvements can be made to extend the versatility of this stable expression strategy, combined with the simple and efficient MACS method. This strategy will then allow a genetic manipulation of target cells/tissues in a wide range of cell types and species. Moreover, this strategy is perfectly suited for primary cells, which are non-immortalized and often delicate, and, more generally, in all cases when time becomes a critical parameter.