Stable integration of an optimized inducible promoter system enables spatiotemporal control of gene expression throughout avian development

Precisely altering gene expression is critical for understanding molecular processes of embryogenesis. Although some tools exist for transgene misexpression in developing chick embryos, we have refined and advanced them by simplifying and optimizing constructs for spatiotemporal control. To maintain expression over the entire course of embryonic development we use an enhanced piggyBac transposon system that efficiently integrates sequences into the host genome. We also incorporate a DNA targeting sequence to direct plasmid translocation into the nucleus and a D4Z4 insulator sequence to prevent epigenetic silencing. We designed these constructs to minimize their size and maximize cellular uptake, and to simplify usage by placing all of the integrating sequences on a single plasmid. Following electroporation of stage HH8.5 embryos, our tetracycline-inducible promoter construct produces robust transgene expression in the presence of doxycycline at any point during embryonic development in ovo or in culture. Moreover, expression levels can be modulated by titrating doxycycline concentrations and spatial control can be achieved using beads or gels. Thus, we have generated a novel, sensitive, tunable, and stable inducible-promoter system for high-resolution gene manipulation in vivo.


Introduction
For thousands of years, birds have been employed to understand how embryos grow, with some of the earliest documented observations made in ancient Greece by Hippocrates (circa 430 BCE) and especially by Aristotle (circa 350 BCE), who methodically recorded detailed empirical descriptions of developing chickens at various stages in the egg (Needham, 1934;Kember, 1971;Churchill, 1991;Aristotle and Thompson, 2001). Such early work not only helped establish developmental biology as a discipline, but also illustrates the fact that artificial incubation and widespread access to embryos have always been a highly advantageous asset of avian model systems (Lillie, 1952;Romanoff, 1960;Hamilton, 1965;Le Douarin and McLaren, 1984;Stern, 1994;Mason, 1999a;Stern, 2005;Jheon and Schneider, 2009).
However, there are limitations to what can be done with avian embryos. First, the initial stages of development from the time of fertilization to the emergence of the blastoderm all happen before an egg is laid (Lillie, 1952;Romanoff, 1960;Kochav et al., 1980;Mason, 1999a). Thus, the earliest events are difficult to access and study. Moreover, by the time a fertilized egg is laid the embryo has already divided into around 60,000 cells (Mason, 1999b;Bellairs and Osmond, 2005). Second, avian eggs can be challenging to manipulate due to the size and abundance of the yolk. Third, unlike mouse or zebrafish model systems, birds have limited genetic tools (Hutt, 1949;Jull, 1952;Abbott, 1967) and targeted mutagenesis followed by forward genetics is difficult. While some transgenic chick and quail lines have been generated (McGrew et al., 2004;Chapman et al., 2005;Koo et al., 2006;van de Lavoir et al., 2006a;van de Lavoir et al., 2006b;Sato et al., 2010;Bower et al., 2011;Huss et al., 2015;June Byun et al., 2017;Tsujino et al., 2019), the technical challenges and expense of making transgenics, combined with the logistics of keeping sufficient transgenic flocks has limited the broad application of this approach (Sang, 2006). However, the ability to create genetic mutations through CRISPR/Cas9 technology has already made the prospects of genome engineering much easier in avians (Ahn et al., 2017;Gandhi et al., 2017;Morin et al., 2017;Williams et al., 2018).
Given the challenges of germ line transgenesis, proxies for studying gene function in avian model systems have predominantly involved a range of alternative strategies. For example, transgenes can be delivered efficiently using retroviral vectors (Fekete and Cepko, 1993a;Morgan and Fekete, 1996;Logan and Tabin, 1998;Chen et al., 1999;Kardon et al., 2003;Hughes, 2004) especially the replication-competent RCAS and RCASBP retroviruses. Some advantages of these vectors include their ability to spread widely throughout host tissues, which in turn allows for broad misexpression of a given transgene, and the ease at preparing large quantities of high-titer viral stocks (Logan and Tabin, 1998). But some limitations of retroviral vectors include the size of the gene insert that they can carry (up to approximately 2.4 kb), as well as their inability to infect most strains of chickens and other birds because of immunity arising from prior exposure to avian sarcoma-leukosis viruses (Hughes, 2004). A further draw-back of retroviral-based strategies is their general lack of precise control over the timing, spatial domains, and levels of gene misexpression. Oftentimes, to achieve sufficient amounts of viral spread, infection must be performed at very early stages, which means that the transgene has to be expressed continuously throughout development regardless if there is a specific stage desired for expression gene of interest.
Another approach for transiently misexpressing genes in a given location or for a certain period of time relies on electroporation of promoter-driven DNA constructs.
Several DNA constructs containing a robust chicken β-actin promoter, a CMV promoter, an internal ribosome entry site (IRES), and a bicistronic reporter with green fluorescent protein (GFP) have been widely adopted including pMES, pCIG, and pCAβ (Swartz et al., 2001;Megason and McMahon, 2002;McLarren et al., 2003;Sauka-Spengler and Barembaum, 2008;Jhingory et al., 2010;Hall et al., 2014;Yang et al., 2014;Gammill et al., 2019;Wu and Taneyhill, 2019). Electroporation can also efficiently enable gene repression using RNA interference (RNAi) and antisense morpholino oligonucleotides (Tucker, 2001;Kos et al., 2003;Chesnutt and Niswander, 2004;Krull, 2004;Nakamura et al., 2004;Rao et al., 2004;Das et al., 2006;Sauka-Spengler and Barembaum, 2008;Gammill et al., 2019). However, due to the extrachromosomal nature of these vectors such treatments are only transient since plasmids and short oligonucleotides degrade and dilute following the proliferation of transfected cells, and misexpression is almost entirely eliminated by 72 to 96 hours (Sauka-Spengler and Barembaum, 2008;Wang et al., 2011;Hall et al., 2014;Bourgeois et al., 2015). Moreover, the promoters in these widely used plasmids cannot be induced to control the timing or levels of gene expression. Thus, there has remained a need for highly versatile vectors that can achieve both long-term and conditional expression in avian embryos. To this end, one transgene expression system was created that uses Tol2 transposon-mediated gene transfer (Koga et al., 1996) to enable stable integration of a given transgene into the avian genome (Kawakami, 2007), and that leverages a tetracycline (tet)-dependent inducible promoter (Sato et al., 2007;Watanabe et al., 2007;Takahashi et al., 2008).
This system has been useful, for example, for studying the behavior and activity of neural crest mesenchyme (NCM) during later stages of embryogenesis (Yokota et al., 2011).
Building on the clear advantages of inducible promoter systems for exerting spatiotemporal control over gene expression and the ability of transposable elements to integrate into the avian genome and facilitate long-term expression throughout development (Wang et al., 2011;Macdonald et al., 2012;Serralbo et al., 2013;Bourgeois et al., 2015), we have designed a new gene delivery system that represents a substantial advance of this technology. Our goal was to streamline and minimize the number of components, to optimize the delivery and detection features, and to achieve efficient and more robust transgene expression. To do so, we generated a constitutively active mNeongreen (GFP) (Shaner et al., 2013) and doxycyline (dox)-inducible (Gossen et al., 1995;Loew et al., 2010;Heinz et al., 2011) mScarlet-I (RFP) (Bindels et al., 2017) construct. Then, to maintain expression of our electroporated constructs over the entire course of embryonic development, we combined our dox-inducible system with an enhanced piggyBac transposon system, which allows for stable semi-random integration into the host genome so that the construct is replicated along with the host genome (Lacoste et al., 2009;Lu et al., 2009;Yusa et al., 2011;Liu et al., 2013;Yusa, 2015). We further improved this construct by adding a D4Z4 genetic insulator sequence to block transcriptional repression (Bire et al., 2013) and a DNA targeting sequence (DTS) to direct transport of the plasmids into the nucleus (Dean et al., 1999;Bai et al., 2017). We find that this construct is sensitive to induction by dox both in ovo and in culture, integrates stably into the genome of chick and duck, and enables expression in embryonic tissues at any desired time or place. Here we demonstrate for example, that NCM can be electroporated at embryonic stage (HH) 8.5 and then gene expression can be induced at HH15, HH30, or later. Also, we show that transgene expression levels can be modulated by titrating the concentration of dox, and precise spatial control over transgene activation can be achieved by implanting beads or gels that release dox locally. Thus, our optimized and integrating inducible-promoter system can control the timing, spatial domains, and levels of gene misexpression throughout avian development, which will be useful for a broad range of experimental contexts.

Design of the small plasmid pNano
To maximize transfection and electroporation efficiency we aimed to generate plasmids as small as possible. Smaller plasmids have been shown to transfect and electroporate more efficiently than large plasmids (Yin et al., 2005). Moreover, large plasmids have been found to be toxic when introduced into cells independent of transgene expression from the plasmid (Lesueur et al., 2016). To minimize the size of our constructs we generated a new plasmid, pNano containing a minimal pMB1 plasmid origin of replication and β-lactamase resistance (BlaR) sequence with a small multicloning site containing EcoRI, EcoRV, and XhoI restriction enzyme sites. The plasmid is 1562 bp ( Fig. 1A) and serves as the backbone for the other constructs generated. To our knowledge, pNano is the smallest plasmid with BlaR selection.

Choosing a promoter
We chose the PGK1 promoter over other commonly used promoters for several reasons. The PGK1 promoter is relatively small at 500 bp and gives quite consistent expression across different cell types (Qin et al., 2010;Huss et al., 2015). The PGK1 promoter does not contain any viral sequences which are prone to epigenetic silencing and loss of expression over time (Brooks et al., 2004;Xia et al., 2007;Norrman et al., 2010).

Choosing a transposon
Transient transfections and electroporations with standard plasmids only enable transgene overexpression for up to 5 days, which is much shorter than the time required to span in ovo development (e.g., 21 days for chick and 28 days for duck). To ensure stable and robust expression over the course of embryogenesis, we used a type II transposon (cut and paste) system to integrate sequences into the genome (Curcio and Derbyshire, 2003;Hickman et al., 2010;Yuan and Wessler, 2011). Several transposable systems currently exist including Tol2 (Koga et al., 1996;Kawakami, 2007), Sleeping Beauty (Ivics et al., 1997), and piggyBac (Fraser et al., 1983;Fraser et al., 1996;Ding et al., 2005). We chose piggyBac due to its several advantages over other transposon systems. Most importantly piggyBac has been shown to have higher transposition activity than Tol2 or Sleeping Beauty in human and chick (Wu et al., 2006;Lu et al., 2009;Huang et al., 2010) and there are improved versions of both the piggyBac transposon and transposase (Lacoste et al., 2009;Yusa et al., 2011). The efficiency of piggyBac integration is relatively size independent up to at least 10 kb (Ding et al., 2005) and piggyBac can deliver cargos in the hundreds of kb (Li et al., 2011;Rostovskaya et al., 2013), while Sleeping Beauty has reduced integration efficiency with cargo sizes above 5 kb (Geurts et al., 2003). PiggyBac semi-randomly integrates into genomes at sites of open chromatin while Sleeping Beauty's integration pattern appears more random (Huang et al., 2010). Successful transposition events into silenced or heterochromatic regions may show no transgene expression due to epigenetic silencing. PiggyBac has lower rates of transgene silencing than Sleeping Beauty or Tol2 ). The piggyBac system is also relatively insensitive to the ratio of transposon to transposase while Sleeping Beauty and Tol2 require titration to determine the optimal ratios . PiggyBac has consistent transposition activity across different cell lines (Wu et al., 2006) and has been utilized in many different organisms including yeast, mice, rats, humans, goat, pig, macaque, chick, rice, and several species of protists and insects (Yusa, 2015). This allows for the same construct to be used among different organisms compared to viral methods which have species-specificity.

Generating the pEPIC1.1 construct for constitutive expression
To enable long term constitutive transgene expression, we generated pEPIC1.1 (enhanced piggyBac IRES mClover3) (Fig. 1B). This construct drives transgene expression under the constitutive PGK promoter. To improve translational efficiency, we included a Kozak sequence directly upstream of the translational start site (Kozak, 1986). As a marker for expression, we used a minimal encephalomyocarditis virus internal ribosome entry site (IRES) (Bochkov and Palmenberg, 2006) to express a bicistronic transcript containing the overexpressed transgene and mClover3 (GFP) (Bajar et al., 2016). An optional C-terminal tandem affinity purification (TAP) tag consisting of 3xFLAG peptide sequences and 2xStrep-tag II sequences (Dalvai et al., 2015) can be added to enhance detection or pulldown. Sequences can be cloned either untagged by digesting pEPIC1.1 with AflII and EcoRI or tagged by digesting with AflII and XhoI. Sustained expression over long time periods is maintained by flanking the overexpression cassette with piggyBac inverted terminal repeat sequences (ITR). The ITRs in the presence of piggyBac transposase (PBase) semi-randomly integrates into the host genome at sequences containing a TTAA motif through a cut and paste mechanism. We used the enhanced piggyBac sequence which contains two point mutations in the left '5 ITR that increase transposition efficiency (Lacoste et al., 2009).
To confirm stable expression, we transfected DF-1 cells with pEPIC1.1 with or without pNano-hyPBase. Following transfection, cells were allowed to express GFP for 2 days to determine the baseline transfection efficiency. We then passaged the cells every 2 days for 20 days, to determine the stability of expression. We found that cells transfected without pNano-hyPBase rapidly lost GFP expression while those transfected with pNano-hyPBase initially had a small drop in GFP expression which then stabilized over time. At 6 days post transfection, cells with pNano-hyPBase retained higher levels of GFP expression compared to those without pNano-hyPBase (75% ±5 compared to 35% ±6, respectively, p < 0.05) ( Fig. 1G-H). By day 20, 70% ±6 of cells transfected with pNano-hyPBase still expressed GFP, compared to < 1% of cells without pNano-hyPBase.
We next confirmed that the pEPIC1.1 construct is functional at the tissue level.
When cells are transfected or electroporated with plasmids, transport from the cytoplasm to the nucleus is required for both expression and transposition into the genome. Plasmid entry into the nucleus generally occurs either during mitosis when the nuclear envelope breaks down, allowing for passive diffusion of plasmids into the nuclear space, or when the intracellular plasmid concentration is very high (10 4 -10 6 molecules of plasmid DNA per cell) (Utvik et al., 1999;Young et al., 2003;Bai et al., 2017). To overcome potential nuclear import barriers, we added a DTS (Dean et al., 1999;Dean et al., 2005). DTSs function by binding to transcription factors which are then actively transported into the nucleus. We chose to use the simian virus 40 (SV40) 72 bp promoter DTS (Dean et al., 1999) because it can function in a wide variety of cell types (Dean, 1997;Young et al., 2003), is small, uses endogenously expressed transcription factors (Miller et al., 2009), and does not require expression of viral proteins (Dean et al., 2005). Alternatively, if nuclear entry is low even with a DTS, addition of trans-cyclohexane-1,2-diol reversibly increases the permeability of the nuclear pore complex allowing plasmids to diffuse into the nucleus (Vandenbroucke et al., 2007;De la Rossa and Jabaudon, 2015;Cervia et al., 2018).
Epigenetic and heterochromatic silencing of foreign DNA inserted into the host genome represent an obstacle for efficient transgene expression both at the time of insertion and over long-term expression (Garrison et al., 2007). Genomic insertions containing viral sequences are known to be actively silenced (Pannell and Ellis, 2001;Ellis, 2005;Wen et al., 2014;Hudecek et al., 2017). An insertion in a heterochromatic region or region that subsequently becomes heterochromatic may result in transgene inactivation (Janssen et al., 2018). To prevent this epigenetic silencing, we added a genetic insulator that blocks the spread of repressive epigenetic marks and heterochromatin (Ali et al., 2016). Moreover, insulator sequences help to protect endogenous sequences from epigenetic activation or silencing caused by the transposition (Hollister and Gaut, 2009). We used the D4Z4 insulator, which is only 65 bp and has been shown to efficiently protect piggyBac transgene expression (Ottaviani et al., 2009;Bire et al., 2013). pPID2 contains two D4Z4 insulator sequences contained within the piggyBac ITRs flanking the multicloning site ( Fig. 2A).

Generating the pPIDNB doxycycline-inducible system
We also added a dox-inducible component to our overexpression constructs, which provides several advantages, including increased temporal control of expression.
Without such precise temporal control, the premature and continuous expression of a gene of interest may disrupt development in ways that cause phenotypes unrelated to the processes under study. A dox-based strategy has several advantages over other inducible systems in that dox is extremely cheap and effective at low concentrations.
Additionally, dox is able to diffuse efficiently through tissues allowing for induction past the surface level (Agwuh and MacGowan, 2006;Sato et al., 2007) and the use of doxsoaked beads or gels can allow for spatial control of expression.
We generated the pPIDNB (piggyBac, insulator, DTS, mNeongreen, bi-directional) construct as a minimal dox-inducible plasmid (Fig. 2B). This plasmid is based upon the pPID2 backbone and includes the DTS, insulator, and piggyBac sequences. In addition, pPIDNB constitutively expresses the reverse tetracycline (tet) transactivator (rtTA) and mNeongreen (GFP) under the PGK promoter (Shaner et al., 2013). The rtTA and mNeongreen coding sequences are bicistronic and are separated by a porcine teschovirus-1 2A (P2A) site which causes them to be expressed as two different peptide sequences (Szymczak et al., 2004;Kim et al., 2011). When bound to dox, the rtTA undergoes a conformational shift allowing binding and activation of the bidirectional tet promoter (Gossen et al., 1995;Das et al., 2016). We chose to use the rtTA-V16 variant of rtTA, which is both sensitive to dox and can induce strong expression (Das et al., 2016). On one side of the bidirectional promoter is mScarlet-I (RFP) serving as a marker for dox induction (Bindels et al., 2017). On the other side of the bidirectional promoter is the cloning site containing AflII and PsiI sites for dox-inducible expression of the gene of interest. Combining both the rtTA and tet promoter into a single construct enables stable inducible-expression with one integrating plasmid and one transposase-expressing plasmid. Moreover, for experiments that would benefit from the ability to detect nuclear localization, we also generated pPIDNB2, which has histone H2B fused to GFP to label nuclei, in contrast to the pPIDNB plasmid where GFP localization is diffuse throughout the cell (Fig. 2C).
To evaluate the sensitivity of the pPIDNB plasmid to induction by dox, we transfected DF-1 cells and performed a dose-response analysis with dox for 24 hours. In the absence of dox, there was a very low basal level of RFP expression, with only 0.15% ±0.2% of the GFP positive cells also expressing detectable levels of RFP expression ( Fig. 2D-E). After treating cells with 2.5 ng/ml dox, 52% ±1.1% of the GFP positive cells also expressed RFP. We found that the percent of RFP expressing cells as a fraction of the GFP positive cells maxed out at a dose of 10 ng/ml dox at 88% ±2.7% with cells treated at 50 ng/ml and 250 ng/ml dox expressing RFP at 80% ±2.7% and 84% ±7.3%, respectively (Fig. 2E). While the fraction of cells expressing RFP did not increase at dox concentrations greater that 10 ng/ml, the intensity of RFP did increase with higher concentrations of dox (Fig. 2D).
We next tested the ability of pPIDNB to drive exogenous gene expression by cloning in the coding sequences for Cxcl14, Gas1 (a plasma membrane receptor, 945 bp), Runx2, and Mmp13. We first transfected DF-1 cells with pPIDNB-Cxcl14, treated with various doses of dox, and found that Cxcl14 expression correlated with the concentration of dox ( Fig 2F). These results along with the RFP data above indicate that dox dose-response is tunable at the cellular level and not simply a binary response to increased dox concentrations causing more cells to express RFP. We found DF-1 cells treated with 2.5, 10, 50, and 250 ng/ml dox for 24 hours increased Cxcl14 mRNA expression by 26 ±14, 96 ±44, 150±70, and 160 ±55 times respectively, compared to cells not treated with dox (Fig. 2F). WB analysis also showed a dose response with 2.5, 10, 50, and 250 ng/ml dox with CXCL14 protein levels increasing by 5.2, 13 ±3.1, 16 ±1.2 and 17 ±2.2 times respectively, compared to cells not treated with dox (Fig. 2G).

Spatiotemporal control of expression in cell culture
To confirm that we could exert spatiotemporal control over transgene expression using pPIDNB, DF-1 cells were transfected with pNano-hyPBase and either pPIDNB-Gas1 or pPIDNB2-Gas1. Cells were passaged for 4 weeks and then sorted for GFP to generate stable lines with either pPIDNB-Gas1 or pPIDNB2-Gas1 integrated into their genomes.
pPIDNB-Gas1 cells showed GFP localized throughout the entire cell while pPIDNB2-Gas1 showed nuclear localization of GFP (Fig. 3A-B). Cells were then treated with 50 ng/ml dox and imaged at 0, 6, and 12 hours post dox treatment. After 6 hours of dox treatment, cells began to express detectable levels of RFP and by 12 hours the RFP signal was robust.
To determine if we could control the spatial localization of transgene expression, we applied minocycline microspheres to DF-1 cells transfected with pPIDNB-Gas1. These microspheres slowly release minocycline, a tetracycline (dox) analog, and induce the tet expression system (Chtarto et al., 2003;Zhou et al., 2006). We applied minocycline microspheres directly to a localized area in the well and cells were imaged at 0, 6, and 12 hours after treatment. After 6 hours, we observed low levels of RFP expression, and after 12 hours RFP expression levels were high in areas adjacent to the microspheres but not in areas further away (Fig. 3C).
For experiments that could benefit from the ability to monitor dynamic changes in the cell cycle, we added a DNA helicase B (DHB) cell cycle sensor sequence (Spencer et al., 2013;Kohrman et al., 2019) to the dox-inducible RFP of pPIDNB2. The DHB cell cycle sensor translocates to the nucleus at G0/G1. During S phase, DHB localizes to both the nucleus and the cytoplasm and during M-phase DHB primarily localizes to the cytoplasm. The nuclear localization of GFP in pPIDNB2 allows for the determination of how much of the DHB signal is nuclear versus cytoplasmic. We transfected DF-1 cells with pPIDNB2 DHB and treated them with 50 ng/ml of dox and imaged them after 12 hours. We found that we could identify cells in different phases of the cell cycle with nuclear-localized DHB (G0/G1), nuclear-and cytoplasm-localized DHB (S phase), and cytoplasm localized DHB (M phase) (Fig. 3D).

Temporal and spatial control of gene expression during development
To exert spatiotemporal control over gene expression in embryonic tissues, we unilaterally electroporated the premigratory cephalic NCM of HH8.5 chick embryos with pPIDNB and pNano-hyPBase. At HH10, we assayed for the extent of electroporation by visualizing GFP-positive cells in ovo in migrating NCM destined for the mandibular primordia (Fig. 4A). These embryos were then incubated until HH30, at which point the mandibular primordia were dissected out, cultured with 50 ng/ml of dox, and imaged at 0, 12, and 24 hours post-treatment. As evidence of the stable genomic integration and induction of the plasmids in embryos, we observe the electroporated side of the mandible expressing GFP, with the contralateral side showing little to no GFP expression. After 12 hours, treatment with dox results in strong RFP signal that is colocalized with GFP and this RFP expression intensifies further by 24 hours (Fig. 4A, Movie 1).
Additionally, some duck embryos were bilaterally electroporated at HH8.5 with pPIDNB-Gas1 and pNano-hyPBase and were treated with 50 ng/ml dox in ovo at HH15. By HH24, we observed RFP expression throughout the mandibular primordia (Fig. 4B). To confirm that in ovo dox treatment would work efficiently even during later stages of development, some chick embryos were unilaterally electroporated at HH8.5 with pPIDNB and pNano-hyPBase, incubated for 7 days, and then were treated in ovo with a single dose of dox (50 ng/ml). These embryos were then allowed to develop for 9 more days (to around HH40), at which point we observed robust unilateral RFP expression in the lower jaw (Fig. 4B).
To exert more precise spatial control over gene expression, some embryos were bilaterally electroporated at HH8.5 with pPIDNB-Gas1 and pNano-hyPBase and incubated until HH40. Their lower jaws were then harvested and either injected along the right side with minocycline microspheres or with dox gel (Pluronic F-127). Pluronic F-127 is a liquid at low temperatures (4 °C) but solidifies at higher temperatures (37 °C) and has been used for delivering drugs to different tissues (Harris et al., 2004;Giovagnoli et al., 2010). After 12 hours of treatment with either minocycline microspheres or with dox gel, we observe RFP expression localized on the right side of the jaw, which becomes more elevated by 24 hours (Fig. 4C-D).

Conclusion
In this study we generated an "all-in-one" piggyBac dox-inducible system. The pPIDNB plasmid is designed to be as small as possible to optimize cellular uptake while incorporating critical features to maximize its functionality. The DTS and insulator sequences serve to promote expression by directing nuclear entry of the plasmid and block heterochromatic silencing expression. We used mutated piggyBac and hyPBase sequences to increase genome integration efficiency. We have also incorporated a constitutively expressed GFP to mark cells that have taken in plasmid DNA and RFP to mark dox-induced cells. Our system facilitates precise temporal control of gene induction and is easily adapted for in vitro or in ovo. Spatial control of gene expression can be achieved by electroporating regions of interest and/or by applying beads or gels to localize the distribution of dox.
The pPIDNB system is able to induce expression quickly and its reliance on a low dose of dox is important because dox has biological effects beyond antimicrobial activity including affecting matrix metalloproteinase activity, inflammation, the NF-κB pathway, and the nervous system (Bahrami et al., 2012;Alexander-Savino et al., 2016). High concentrations of dox (e.g., 1000 ng/ml) are cytotoxic in culture and have strong proliferative and metabolic effects, and some cell types are affected at even lower concentrations (e.g., 100-200 ng/ml) (Ermak et al., 2003;Ahler et al., 2013;Alexander-Savino et al., 2016). By using a low dose of dox (50 ng/ml) we have likely minimized any off-target effects of dox treatment.
While in the current study, we designed the pPIDNB construct for transgene overexpression, we envision that future applications will include different types of experiments such as gene knockdown using CRISPRi (Qi et al., 2013;Mandegar et al., 2016). For example, catalytically inactive Cas9 could be placed with transcriptional repressors under an inducible tet promoter (Qi et al., 2013;Yeo et al., 2018).
Constitutively active U6 promoters would drive expression of single guide RNAs (Cong et al., 2013;Gandhi et al., 2017;Williams et al., 2018). Using similar protocols for overexpression or knockdown would reduce the number of variables between experiments and help limit the confounding effects from different constructs. Overall, a great strength of avian model systems has been the combination of experimental embryology and modern genetic techniques. Our new, sensitive, stable, and robust inducible-promoter system builds on this strength and joins an arsenal of tools for manipulating gene expression that will likely be useful to the broader developmental biology community.

Plasmids
To generate pNano, the Ori and BlaR from pJet1.2 (ThermoFisher, K1231) were  Hamilton (HH) staging system, a well-established standard which utilizes an approach based on external morphological characters and that is independent of body size and incubation time (Hamburger and Hamilton, 1951;Hamilton, 1965;Ricklefs and Starck, 1998;Starck and Ricklefs, 1998). For late embryonic stages, we relied primarily on growth of the limbs, facial primordia, feather buds, and eyes (Eames and Schneider, 2005;Eames and Schneider, 2008;Merrill et al., 2008).
In ovo electroporations were performed using a solution of pPIDNB and pNano-hyPBase at 3 µg/µl and 1 µg/µl, respectively. With the addition of Fast Green tracer dye, DNA solution was injected into HH8.5 chick neural tubes with a Pneumatic PicoPump using thin wall borosilicate glass micropipettes pulled on a micropipette puller. Platinum electrodes were positioned on each side of the area pellucida, centered on the midbrain-hindbrain boundary. For unilateral electroporations, we delivered three square pulses at 50 V for 1 ms spaced 50 ms apart followed by five square pluses at 10 V for 50 ms spaced 50 ms apart. For bilateral electroporations, we delivered three square pulses at 50 V for 1 ms spaced 50 ms apart, three square pulses at 50 V for 1 ms spaced 50 ms apart in the reverse polarity, three five square pluses at 10 V for 50 ms spaced 50 ms apart followed by, five square pluses at 10 V for 50 ms spaced 50 ms apart in the reverse polarity.
qPCR DNased RNA was reverse-transcribed using iSCRIPT (Bio-Rad, 1708841). Gene expression was quantified by qPCR with iQ SYBR Green Supermix (Bio-Rad, 1708882) and normalized to 18S rRNA following previously published protocols (Dole et al., 2015;Smith et al., 2016). Primer sets were designed and optimized as described previously (Ealba and Schneider, 2013) and are listed in supplemental Table S1. Each sample was assayed in technical duplicate.
Fluorescent signal was captured using the Odyssey Imaging System (ThermoFisher).
Quantifications of protein bands were performed using Image Studio Lite. RUNX2, MMP13, and CXCL14 levels were normalized to β-actin.

Doxycycline treatment
Stock solutions of doxycycline hyclate (Acros Organics, 446060250) were made to a final concentration of 1 mg/ml in water, filter sterilized, and stored at -20 °C as single use aliquots. DF-1 cells and mandibles were treated in culture with the stock solution diluted in DMEM, with minocycline microspheres (Arrestin) added directly to each well, or by suspending microspheres in PBS and injecting them into the lower jaw with a 30gauge needle. Pluronic F-127 (Sigma, P2443-250G) was dissolved at a final concentration of 35% (w/v) in DMEM growth medium rocking at 4°C for 48 h. Dox was added to Pluronic F-127 for a final concentration of 500 ng/ml and injected into the lower jaw with an 18-gauge needle. For in ovo treatments, 2.5 µl (for chick) and 3.75 ul (for duck) of the 1mg/ml dox stock solution was diluted with 750 µl of HBSS. This solution was then gently pipetted into the egg adjacent to the embryo and allowed to diffuse.

Fluorescence-activated cell sorting (FACS)
DF-1 cells were washed with 2 mL of Trypsin followed by 3mL fresh wash. Trypsin activity was inhibited by adding 5 mL of DMEM with 10% FBS. Cells were pipetted and passed through 70 µm filter. Cells were sorted on FACSAriaII Flow Cytometer (BD Bioscience, San Jose, CA). For all sorts, debris and dead cells were eliminated using FSC-A and SSC-A gating, doublets were excluded via gating discrimination using FSC-H and FSC-W, and only GFP+ cells were collected.

Statistical analysis
Statistical analysis carried out using Student's t-test was performed (GraphPad Prism version 8.4.3,GraphPad Software,La Jolla,CA). When multiple comparisons were made, p-values were adjusted using the Benjamini-Hochberg procedure (Benjamini and Hochberg, 1995).
for fertilized chick and duck eggs; and the UCSF Biological Imaging Development Core (BIDC) for microscopy support. The pmScarlet-i_C1 was a gift from Dorus Gadella (Addgene, #85044). The AAVS1 Puro Tet3G 3xFLAG Twin Strep was a gift from Yannick Doyon (Addgene, # 92099). The pKanCMV-mClover3-mRuby3 was a gift from Michael Lin (Addgene, #74252). The DHB was a gift from Dave Matus. The PGK promoter was a gift from Jonathan Brunger via Tamara Alliston.

Competing Interests
The authors declare no competing interests.

Funding
This work was supported in part by National Institutes of Health F30 DE027616 to A.N.; F31 DE027283 to S.S.S.; and R01 DE016402, R01 DE025668, and S10 OD021664 to R.A.S.

Data Availability
All datasets and constructs will be made publicly available at the time of publication.  Cephalic neural crest mesenchyme (NCM) electroporated unilaterally with pPIDNB and pNano-hyPBase in a chick embryo at HH8.5 constitutively expresses mNeonGreen (GFP) as shown at HH10 (counterstained red with neutral red). At HH30, the lower jaw shows unilateral GFP expression in NCM. After 12 and 24 hours in culture, NCM express mScarlet (RFP) in response to treatment with 50 ng/ml doxycycline (dox). B) NCM bilaterally electroporated with pPIDNB and pNano-hyPBase in a duck embryo at HH8.5 shows RFP expression on both sides of the lower jaw at HH24 (after bilateral electroporation) when treated in ovo with 50 ng/ml dox at HH15. NCM bilaterally electroporated with pPIDNB and pNano-hyPBase in a chick embryo at HH8.5 shows RFP expression on one side of the lower jaw at HH40 (after unilateral electroporation) when treated in ovo with 50 ng/ml dox 7 days after electroporation and imaged 9 days after treatment. C) NCM electroporated bilaterally with pPIDNB-Gas1 and pNano-hyPBase in a duck embryo at HH8.5 shows RFP expression in the lower jaw (white dashed area) 12 and 24 hours after being injected in culture with minocycline microspheres at HH40. D) NCM electroporated unilaterally with pPIDNB-Gas1 and pNano-hyPBase in a duck embryo at HH8.5 shows RFP expression in the lower jaw (white dashed area) 12 and 24 hours after being treated at HH40 with 35% Pluronic F-127 gel containing 50 ng/ml dox (white asterisk).