A Hox complex activates and potentiates the Epidermal Growth Factor signaling pathway to specify Drosophila oenocytes

Hox transcription factors specify distinct cell types along the anterior-posterior axis of metazoans by regulating target genes that modulate signaling pathways. A well-established example is the induction of Epidermal Growth Factor (EGF) signaling by an Abdominal-A (Abd-A) Hox complex during the specification of Drosophila hepatocyte-like cells (oenocytes). Previous studies revealed that Abd-A is non-cell autonomously required to promote oenocyte fate by directly activating a gene (rhomboid) that triggers EGF secretion from sensory organ precursor (SOP) cells. Neighboring cells that receive the EGF signal initiate a largely unknown pathway to promote oenocyte fate. Here, we show that Abd-A also plays a cell autonomous role in inducing oenocyte fate by activating the expression of the Pointed-P1 (PntP1) ETS transcription factor downstream of EGF signaling. Genetic studies demonstrate that both PntP1 and PntP2 are required for oenocyte specification. Moreover, we found that PntP1 contains a conserved enhancer (PntP1OE) that is activated in oenocyte precursor cells by EGF signaling via direct regulation by the Pnt transcription factors as well as a transcription factor complex consisting of Abd-A, Extradenticle, and Homothorax. Our findings demonstrate that the same Abd-A Hox complex required for sending the EGF signal from SOP cells, enhances the competency of receiving cells to select oenocyte cell fate by up-regulating PntP1. Since PntP1 is a downstream effector of EGF signaling, these findings provide insight into how a Hox factor can both trigger and potentiate the EGF signal to promote an essential cell fate along the body plan.


Introduction
Hox genes encode a family of conserved homeodomain transcription factors that pattern the developing anterior-posterior (A-P) axis of metazoans by regulating target gene expression [1][2][3][4][5]. Most animals contain five or more Hox genes with each being expressed in distinct A-P regions (i.e. segments) to specify distinct morphological structures and cell fates in organs such as the nervous system, musculature, digestive system, and appendages [6,7]. Developmental genetic and genomic studies have revealed that Hox factors interact with additional morphogenetic pathways to specify diverse cell types via the regulation of numerous downstream target genes [8][9][10][11]. Ultimately, these target genes contribute to gene regulatory networks (GRNs) that ensure appropriate cell fate specification within different organ systems [12]. Understanding how Hox factors intersect with other growth control pathways is therefore likely to reveal fundamental insight into both animal development and evolution.
Functional studies in organisms from worms to mice have revealed that the integration of Hox factors with signaling pathways is a common mechanism used for the specification of distinct cell types. Examples include the integration of Wnt, Epidermal Growth Factor (EGF) and Notch signals with Hox factors during the specification of different epidermal and vulval fates in C. elegans [13,14], the integration of retinoic acid (RA) and Fibroblast Growth Factor (FGF) signaling with Hox factors during spinal cord and hindbrain development in vertebrates [15][16][17][18], and the integration of the Jak/Stat signaling pathway with the Abdominal-B (Abd-B) Hox factor during the formation of the Drosophila genitalia [19,20]. While the gene regulatory networks underlying Hox-mediated cell type specification are just beginning to be defined, these studies emphasize that synergy between Hox factors and growth factor signaling pathways are frequently required for proper organogenesis.
The development of larval oenocytes in the Drosophila embryo represents a well-characterized example of how a Hox factor (Abdominal-A, Abd-A) specifies a segment-specific cell type by directly regulating a signaling pathway [21,22]. Larval oenocytes are hepatocyte-like cells that regulate lipid and nutrient metabolism required for animal growth [23]. Together with the cells of the fat body, larval oenocytes perform many of the functions of the mammalian liver. Larval oenocytes are specified in clusters of three to nine cells from the dorsal ectoderm of seven abdominal segments (A1 through A7) during Drosophila embryogenesis [22]. Developmental genetic studies have uncovered that the Epidermal Growth Factor (EGF) pathway, the Spalt transcription factors, and the Abdominal-A (Abd-A) Hox factor as well as the Extradenticle (Exd) and Homothorax (Hth) Hox co-factors are all required for proper oenocyte development [22,[24][25][26]. The current model for oenocyte specification is that an Abd-A/Hth/ Exd transcription factor complex induces the expression of the rhomboid (rho) gene in a specific subset of abdominal sensory organ precursor cells (SOPs) [27][28][29]. rho encodes a serine protease that cleaves and promotes the secretion of the Spitz (Spi) EGF ligand, and the neighboring ectoderm cells of the dorsal ectoderm, which express the Spalt-family of zinc finger transcription factors (Spalt-major (Salm) and Spalt-related (Salr)), receive the EGF ligand, further up-regulate the Spalt transcription factors, and initiate a largely unknown genetic cascade that commits the cell to an oenocyte fate [24,25,30,31]. Other Hox factors fail to activate the rho protease in SOP cells, and thus the EGF signal and subsequent specification of oenocyte fate is restricted to Abd-A-positive segments.
The above model suggests that Abd-A, Exd, and Hth are non-cell autonomously required to specify oenocytes via the activation of rho expression and thereby stimulate EGF secretion to promote larval oenocyte development. Consistent with this idea, previous studies have shown that ectopic stimulation of the EGF pathway in thoracic segments, which do not express Abd-A, is sufficient to induce oenocytes [22]. However, in this study, we show that Abd-A also plays a positive role within oenocyte precursors by potentiating the EGF signal. First, we used genetics to demonstrate that the two ETS transcription factors encoded by pointed (pnt) (pntP1 and pntP2) are each required for oenocyte specification. Both Pnt factors, which are produced via alternative promoters, are activated by EGF signaling albeit via distinct mechanisms [32][33][34][35][36]. PntP2 requires direct phosphorylation by MAPK to activate transcription, whereas PntP1 is transcriptionally up-regulated by the EGF pathway through largely unknown mechanisms [37]. Second, we identified a PntP1 specific cis-regulatory module (CRM) that is activated in oenocytes via direct regulation by an Abd-A/Exd/Hth Hox complex. In addition, we found that this PntP1 enhancer (PntP1OE) is stimulated by the EGF pathway and contains ETS binding sites required for optimal enhancer activation. These findings support a model in which a transient EGF signal phosphorylates and activates the PntP2 transcription factor, and PntP2, in parallel with an Abd-A Hox complex, directly activates PntP1OE transcription in oenocyte precursor cells. Importantly, unlike PntP2, the PntP1 transcription factor is constitutively active and does not require MAPK phosphorylation to activate gene expression [32,36,37]. Thus, Abd-A not only directly stimulates EGF secretion from neighboring abdominal SOP cells, but also potentiates the EGF signal by stimulating PntP1 expression within oenocyte precursor cells.

Results
The Pointed-P1 and Pointed-P2 ETS factors are both required for the specification of larval oenocytes While genetic studies have shown that the EGF signaling pathway is required for larval oenocyte development, much less is known about the downstream effectors of oenocyte specification [21,38]. To determine if the Pnt transcription factors are required for the development of these cells, we first analyzed embryos carrying a deletion within the pnt locus (pnt Δ88 [34]) that removes both pntP1 and pntP2 (Fig 1A). In contrast to wild type or pnt Δ88 heterozygote animals, pnt Δ88 null embryos lack oenocytes (marked by high Spalt-major (Salm) levels, Fig 1B  and 1C). Analysis of embryos containing smaller deletions that only remove pntP1 (pnt Δ33 ) or pntP2 (pnt Δ78 ) [36,39] similarly resulted in a complete loss of oenocytes, demonstrating that both pnt gene products are required to make this cell type (Fig 1D and 1E). Consistent with this finding, we found that both PntP1 and PntP2 are expressed in developing oenocytes. To analyze pntP2 expression, we used a PntP2-lacZ line (P(lacW)pnt 1277 [40]) and found β-gal expression in cells that include the early oenocyte precursors (Fig 1F). Similar analysis using a PntP1-specific antibody [41] revealed that PntP1 was also expressed in the developing oenocytes as marked by SvpΔ18-lacZ, albeit at a slightly later time point than pntP2-lacZ expression (Fig 1G). Thus, both pnt gene products are expressed within and required for the development of larval oenocytes during Drosophila embryogenesis.
To determine if the expression of each Pnt factor is sufficient to specify ectopic oenocytes, we used the PrdG4 driver to express either UAS-PntP1 or UAS-PntP2 in every other segment of the Drosophila embryo. In this assay, only PntP1 was sufficient to induce ectopic oenocytes in PrdG4+ thoracic segments and extra oenocytes (HNF4+ cells) in PrdG4+ abdominal segments, whereas PntP2 had no effect on oenocyte specification (Fig 2A-2C, S1 Fig). Moreover, consistent with Spalt factors being significantly up-regulated by EGF signaling during larval Map of pointed locus showing the alternative promoters and transcripts that produce the pntP1 and pntP2 gene products as well as the deleted regions (green) that generate loss-of-function alleles for pntP1 and pntP2 (pnt Δ88 ), pntP1 (pnt Δ33 ) and pntP2 (pnt Δ78 ) [34,36,39]. The pntP2-lacZ (P(lacW)pnt 1277 [40]) and pnt M103880 insertions are noted and the PntP1OE enhancer is highlighted. Fig 1B-1E: Lateral views of stage 15 Drosophila embryos with the following alleles: control pnt Δ88 heterozygotes (pnt Δ88 /Tm3,ftz-Z) (B), pnt Δ88 homozygotes (C), pnt Δ33 homozygotes (D), and pnt Δ78 homozygotes (E). Embryos were immunostained for Spalt-major (Salm, green) and β-gal (red). The first abdominal segment (A1) is labeled for each embryo. Note clustered oenocytes (marked by high levels of Salm) only form in pnt Δ88 heterozygotes whereas the broader Salm expression pattern in the dorsal ectoderm is observed in all embryos. Fig 1F: Lateral view of pntP2-lacZ Drosophila embryo (stage 11) immunostained for Salm (green) and β-gal (red) reveals pntP2 activity in early oenocyte precursor cells (arrowheads). Fig 1G: Lateral view of SvpΔ18-lacZ Drosophila embryo (late stage 11) immunostained for PntP1 (green) and β -gal (red) reveals PntP1 expression in oenocyte precursor cells (arrowheads) shortly after their initial specification. oenocyte specification [23,24], we found that PntP1 increases Salm expression levels in both thoracic and abdominal dorsal ectoderm cells in early embryos (Fig 2D). Since PrdG4 is active in both the EGF sending (SOP) and receiving cells, we also tested the ability of PntP1 to induce oenocytes using Gal4 drivers expressed only within the SOP lineage (AtoG4) versus the surrounding ectoderm (SpaltG4) [22]. Consistent with PntP1 functioning cell autonomously within oenocyte precursors, we found that only SpaltG4;UAS-PntP1 embryos produced ectopic oenocytes (Fig 2E and 2F). Overall, these findings suggest that the constitutively active PntP1 transcription factor is sufficient to induce oenocyte formation in the dorsal ectoderm, whereas PntP2 likely requires MAPK phosphorylation downstream of EGF signaling to activate gene expression and promote oenocyte cell fate [34,36,37]. Next, we wanted to determine if PntP1 is sufficient to induce oenocyte fate in the absence of EGF signaling. To address this question, we used PrdG4 to express UAS-PntP1 together with a dominant negative EGF receptor molecule (UAS-EGFR DN ). Interestingly, expressing EGFR DN on its own or with PntP1 resulted in a complete suppression of oenocyte specification (Fig 2G and 2H). Altogether, these findings suggest that while PntP1 is required to make oenocytes in the dorsal ectoderm, it is not sufficient to do so in the absence of EGF signaling. Thus, the EGF signaling pathway is required to regulate additional factors besides PntP1 to specify oenocyte fate.

Identification of a PntP1 enhancer that activates gene expression in oenocyte precursor cells
While it has been established that EGF signaling activates the expression of pntP1 at the transcriptional level, the cis-regulatory modules (CRMs) controlling pntP1 expression are largely unknown. We identified a 700 base-pair sequence within the pnt locus (Fig 1A) that is sufficient to activate reporter expression (PntP1OE-lacZ) within oenocytes (Fig 3A and 3B). PntP1OE-lacZ expression is first detected in Salm+ oenocyte precursor cells during their specification, and PntP1OE-lacZ expression is highly enriched within mature HNF4-positive oenocytes of older embryos (Fig 3A and 3B). Interestingly, genomic analysis uncovered a fly line with a Minos transposable element (pnt MI03880 ) inserted within the PntP1OE enhancer of the PntP1OE is an essential enhancer for the specification of oenocytes. Fig 3A: Lateral view of a PntP1OE-lacZ embryo (stage 11) immunostained for β-gal (red) and Salm (green) reveals enhancer activity in larval oenocyte precursor cells. Fig 3B: Lateral view of a PntP1OE-lacZ embryo (stage 15) immunostained for β-gal (red) and HNF4 (green) reveals enhancer activity in oenocytes. Fig 3C and 3D: Lateral view of a pnt MI03880 /TM6B-22uZ heterozygote (C) and pnt MI03880 homozygote (D) embryo (stage 15) immunostained for β-gal (red) and Salm (green) reveals loss of oenocytes in pnt MI03880 mutant embryos. Fig 3E and 3F: Lateral view of pnt MI03880 /pnt Δ78 (E) and pnt MI03880 /pnt Δ33 (F) transheterozygote embryos (stage 15) immunostained for β-gal (red) and Salm (green) reveals that the pnt Δ78 (PntP2) null allele but not the pnt Δ33 (PntP1) null allele complements pnt MI03880 .
https://doi.org/10.1371/journal.pgen.1006910.g003 pnt locus (Fig 1A). Embryos homozygous for this insertion lack oenocytes but retain pntP1 expression in other regions of the Drosophila embryo such as the tracheal pits (Fig 3C and 3D,  S2 Fig). To assess if pnt MI03880 selectively disrupts either pntP1 or pntP2 activity, we generated transheterozygote embryos that revealed the pnt MI03880 insertion complements the pntP2 allele (pnt Δ78 ) but fails to complement the pntP1 allele (pnt Δ33 ) (Fig 3E and 3F). Taken together, these findings strongly suggest that PntP1OE enhancer regulates pntP1 expression in oenocyte precursor cells.

EGF signaling and the Pnt transcription factors regulate PntP1OE enhancer activity in oenocyte precursors
The expression of pntP1 is often induced by receptor tyrosine kinase (RTK) signaling in Drosophila via relatively unknown mechanisms. To determine if PntP1OE is directly regulated by EGF signaling in oenocyte precursors, we first analyzed the PntP1OE sequence and identified seven conserved ETS binding motifs within a 480 base pair region (PntP1OE 480 , Fig 4A). Transgenic reporter analysis of PntP1OE 480 -lacZ reveals it is sufficient to activate gene expression in oenocyte precursors, and mutation of all seven ETS sequences (PntP1OE 480ETS -lacZ) results in a significant loss of reporter activity (Fig 4B-4D 480 and OE 480ETS sequences). These findings are consistent with Pnt transcription factors playing a direct role in enhancer activation downstream of EGF signaling. Congruent with this idea, we found that activation of the EGF pathway via ectopic expression of rho (PrdG4;UAS-rho) or PntP1 (PrdG4;PntP1) was sufficient to induce PntP1OE-lacZ activity (Fig 4E and 4F). In contrast, expression of PntP2 (PrdG4;UAS-PntP2), which requires an EGF-dependent phosphorylation event for activity, did not significantly alter PntP1OE-lacZ activity (Fig 4G).
To determine if both PntP1 and PntP2 are required for PntP1OE-lacZ activation in oenocyte precursors, we next analyzed embryos mutant for both pntP1 and pntP2 (pntΔ88), pntP1 alone (pntΔ33), or pntP2 alone (pntΔ78). In all cases, we observed a dramatic loss of PntP1OE-lacZ expression in older embryos (S4 Fig). However, analysis of PntP1OE-lacZ activity in early embryos revealed that pntP1 mutants expressed weak β-gal levels (Fig 4H and 4I), whereas mutants that lacked pntP2 (either alone or in combination with pntP1) showed no early PntP1OE-lacZ activity (Fig 4J and 4K). These findings are consistent with an EGF signal activating the PntP2 protein via phosphorylation and thereby stimulating the PntP1OE enhancer to induce PntP1 expression. Once activated, PntP1 can then positively auto-regulate itself. Hence, in the absence of pntP2, PntP1OE-lacZ expression is not observed, and no PntP1 protein is produced. In contrast, without pntP1, PntP2 induces an initial up-regulation of PntP1OE-lacZ expression during active EGF signaling, but its expression is not maintained. Moreover, while pntP1 mutants can weakly induce PntP1OE-lacZ expression, these cells fail to significantly up-regulate Spalt expression and develop into oenocytes (compare Fig 4H and  4I). Thus, PntP1 and PntP2 are required to both upregulate PntP1OE and induce oenocytes.

The Abdominal-A Hox factor and Homothorax enhance oenocyte specification downstream of EGF signaling
The current model of larval oenocyte specification is that an Abd-A Hox complex with Extradenticle (Exd) and Homothorax (Hth) is required to induce an EGF signal by direct activation of the rhomboid (rho) serine protease in the abdominal C1 sensory organ precursor (SOP) cells [28,38]. The neighboring cells receive the EGF signal and activate the Pnt transcription factors to initiate the specification of larval oenocyte fate. If Abd-A is only non-cell autonomously required to activate rho to induce oenocyte fate, we reasoned that equal expression of rho in the thorax will bypass the requirement for Abd-A in making oenocytes. To test this idea, we first established that the PrdG4 driver can be used to drive equal expression of Rho (UAS-Rho; PrdG4) and thereby EGFR activation within thoracic and abdominal segments as evidenced by equivalent activity of phospho-ERK (Fig 5A and 5B) [43]. Consistent with previous studies [22], we found that thoracic expression of Rho was sufficient to induce oenocytes, indicating Abd-A is not strictly required for oenocyte development. However, comparisons between thoracic and abdominal segments revealed dramatic differences in the number of oenocytes specified with nearly six times as many made in abdominal segments than in the thorax (Fig 5E  and 5I). Importantly, co-expression of Abd-A with Rho (PrdG4;UAS-Rho;UAS-Abd-A) stimulated equal numbers of oenocytes in both thoracic and abdominal segments (Fig 5F and 5I). Moreover, the expression of Abd-A alone is sufficient to induce the same number of oenocytes  (approximately 6) in the thorax and the abdomen (Fig 5D and 5I). Altogether, these findings suggest that while both thoracic and abdominal cells are equally capable of receiving the EGF signal (equal phospho-Erk levels), the abdominal cells are more highly competent to become oenocytes in an AbdA-dependent manner.
Next, we wanted to assess the role of Hth, a Hox co-factor in oenocyte specification. As mentioned above, Hth forms a direct complex with Abd-A on a rho cis-regulatory module to stimulate gene expression in abdominal SOPs [28,38]. Loss-of-function genetic studies revealed that oenoctyes fail to form in a strong hth hypomorph (hth P2 , Fig 5I) [44]. To determine if the loss of oenocytes is simply due to a failure to activate rho expression and thereby EGF secretion, we used PrdG4 to express Rho in the absence of hth function (UAS-rho;PrdG4, hth P2 ) and found that no oenocytes were induced by high levels of Rho expression in the thorax and very few oenocytes were induced in the abdomen of hth mutant embryos (Fig 5G and  5I). Thus, these findings demonstrate that like Abd-A, Hth function is not only essential for activating rho in SOPs but it also greatly enhances the specification of oenocytes downstream of EGF signaling.

An Abdominal-A Hox complex regulates PntP1OE activity in oenocyte precursor cells
The ability of Abd-A to enhance the competency of cells to commit to an oenocyte fate downstream of EGF signaling suggests that this abdominal Hox factor regulates target genes in the oenocyte specification pathway such as the PntP1OE enhancer. Consistent with this idea, we found that PntP1OE-lacZ expression was induced by ectopic Abd-A (PrdG4;UAS-AbdA) in the thorax (S5 Fig). Moreover, sequence analysis of PntP1OE revealed two regions with conserved binding sites for Hox and the Three Amino-acid Loop Extension (TALE) homeodomain factors that include Hth and Exd (Fig 6A and 6B). Electromobility Shift Assays (EMSAs) using purified Exd, Hth, and Abd-A proteins reveals these proteins bind probes containing each site and point mutations within these sites greatly decrease complex formation (Fig 6C). In contrast, equimolar concentrations of the thoracic Hox factor Antennapedia (Antp) binds these probes with significantly less affinity, consistent with Abd-A selectively enhancing the competency of oenocyte specification (Fig 6D). To determine if these sites are essential for proper PntP1OE-lacZ regulation in oenocytes, we generated transgenes containing Hox/Exd/Hth binding site mutations within both regions and found a significant loss of oenocyte enhancer activity (Fig 6E-6G).
Altogether, these data are consistent with an Abd-A Hox complex enhancing oenocyte competence by potentiating the EGF signal in the signal receiving oenocyte precursor cells. If pntP1 is the only target gene required to be activated by Abd-A to promote oenocyte UAS-Rho embryo (stage 11) immunostained for phospho-ERK. The PrdG4-active second thoracic (T2) and first abdominal (A1) segments are labeled. Fig 5B: Quantitation of phospho-ERK expression in the PrdG4 + thoracic and abdominal segments of PrdG4;UAS-Rho embryos reveals equal activity between segments. Fig 5C-5F: Lateral views of wild type (C), PrdG4;UAS-AbdA (D), PrdG4;UAS-Rho (E), and PrdG4;UAS-Rho; UAS-AbdA (F) embryos (stage 15) immunostained for HNF4 (blue). The PrdG4+ thoracic (T2) and abdominal (A1) segments are labeled. Note, oenocyte numbers in wild type animals can vary from four to ten per abdominal segment [21,24,26]. Quantitative assessment of significant changes in oenocyte numbers are reported in Fig 5I. Fig 5G: Lateral view of hth p2 ;PrdG4;UAS-Rho embryo (stage 15) immunostained for HNF4 (blue). The PrdG4+ thoracic (T2) and abdominal (A1) segments are labeled. formation, we reasoned that ectopic expression of PntP1 using PrdG4 (PrdG4;UAS-PntP1) should induce as many oenocytes in the thorax as the abdomen. However, we found that like rho, PntP1 expression is not sufficient to induce as many oenocytes in thoracic segments that lack Abd-A expression compared to the Abd-A expressing abdominal segments (see Fig 2B  and Fig 5H and 5I). These findings suggest that Abd-A is likely to enhance oenocyte competency by regulating additional downstream target genes and not just through the potentiation of the EGF signal via pntP1 activation.

Discussion
In this study, we investigated how a key component of the anterior-posterior specification pathway (the Abd-A Hox factor) regulates the Epidermal Growth Factor (EGF) signaling pathway to promote formation of hepatocyte-like cells (oenocytes). Prior studies had shown that Abd-A is non-cell autonomously required to promote oenocyte fate by directly activating a gene (rhomboid) that causes the secretion of an EGF ligand (Spitz) from a subset of abdominal sensory organ precursor (SOP) cells [22,27,28]. Here, we show that Abd-A also plays a cell autonomous role in promoting oenocyte fate by directly regulating the transcriptional expression of the PntP1 ETS transcription factor (Fig 7). Since the activation of Pnt transcription factors lies downstream of EGF signaling [35], these findings suggest that Abd-A potentiates the EGF signal to promote oenocyte fate. Below, we discuss how these data provide new insights into both the mechanisms by which the Hox and EGF signaling pathways intersect as well as the gene regulatory networks that are required for oenocyte specification.
The EGF signaling pathway largely regulates the specification of cell types by regulating the expression of downstream target genes via ETS transcription factors [35]. Drosophila has three ETS transcription factors regulated by the EGF pathway. The PntP1 and PntP2 ETS proteins largely function as transcriptional activators that are produced via alternative promoters, whereas the Yan ETS factor functions as a transcriptional repressor to thereby antagonize the EGF pathway [33,34,36,37,45]. Of these three proteins, both Yan and PntP2 are directly modified by the MAPK pathway. PntP2 requires MAPK phosphorylation to activate transcription, while MAPK phosphorylation of Yan results in its nuclear export and degradation. In contrast, the PntP1 ETS factor is regulated by RTK signaling at the transcriptional level, and once expressed, PntP1 does not require post-translational modifications to activate gene expression. Moreover, at least in the Drosophila pupal eye, the half-life of the PntP1 protein can be as long as 6 hours [32]. Thus, EGF signaling transiently promotes target gene expression via post-translational modifications of two ETS factors and the transcriptional induction of PntP1 results in a longer lasting effect on target gene expression.
Consistent with Pnt factors being key effectors of the EGF pathway, we found that both PntP1 and PntP2 are required to specify larval oenocytes in the Drosophila embryo. Together with previous studies showing that loss-of-function yan mutations result in the formation of extra abdominal oenocytes (~12 per cluster compared to~6 per cluster in wild type embryos binding sites are shaded and the sequence mutations tested in EMSAs and quantitative reporter assays are noted. Fig 6C: DNA binding assays using purified Abd-A and Exd/Hth proteins and wild type and mutant Hox/TALE probes. Gray arrows denote Exd/ Hth complex formation on DNA, whereas red arrowheads denote Exd/Hth/Abd-A complex formation on DNA. Note, diminished DNA binding activities on mutant probes. Fig 6D: Comparisons in DNA binding activity and complex formation with Exd/Hth between equimolar concentrations of purified Abd-A (AA) and the Antp (AN) thoracic Hox factor. Gray arrows denote Exd/Hth complex formation on DNA, whereas red arrowheads denote Exd/Hth/Hox complexes on DNA. Note, diminished DNA binding activity of Antp compared to Abd-A. Fig 6E and 6F: Lateral views of PntP1OE-lacZ (G), and PntP1-OE(hox/TALE) 2 -lacZ (H) embryos (stage 15) immunostained for β-gal and HNF4 (blue). Fig 6G: Quantitation of PntP1OE-lacZ and PntP1OE(hox/TALE) 2 -lacZ reporter activity reveals a significant loss of enhancer activity when the Hox/TALE motifs are mutated. * denotes p-value<0.01.
https://doi.org/10.1371/journal.pgen.1006910.g006 [26]), these findings support the following model: Abd-A/Hth/Exd complexes activate the expression of the rhomboid protease to trigger EGF ligand secretion from the C1 abdominal SOP cells [27,28]. The neighboring dorsal ectoderm cells receive the transient EGF signal, stimulate the MAPK signal transduction pathway, and thereby activate PntP2 and inactivate Yan [24,25,45]. Activated PntP2 as well as the Abd-A/Hth/Exd complexes bind the PntP1OE enhancer to activate PntP1 transcription within oenocyte precursors (Fig 7). Once expressed, the constitutively active PntP1 protein can positively auto-regulate itself to further potentiate the EGF signal and promote oenocyte specification. However, it should be noted that if the EGF signal is of sufficient strength, the cell autonomous requirement of the Abd-A Hox factor for oenocyte specification can be bypassed. For example, direct activation of rho in the thorax can induce oenocytes, but to a much lesser extent than in abd-A expressing segments (Fig 5). Taken together with previous publications [22,28], these studies demonstrate that an Abd-A/Hth/Exd transcription factor complex contributes to oenocyte development in two ways: direct activation of the EGF signal within the C1 SOP and direct potentiation of the EGF signal in the receiving cell by activation of pntP1 during oenocyte differentiation (Fig 7).
Two additional findings from our studies suggest that further integration of EGF and Abd-A Hox factors are likely to occur during oenocyte specification beyond the activation of pntP1. First, our studies reveal that EGF signaling must regulate additional factors required for the specification of oenocyte fate as the expression of high levels of PntP1 is not sufficient to induce oenocytes when the EGF signal is inhibited using a dominant negative protein (Fig 2H). Second, Abd-A must also regulate additional factors in oenocyte precursors besides PntP1, as the expression of PntP1 in thoracic segments is unable to induce as many oenocytes compared with abdominal segments that express Abd-A (Fig 5H and 5I). While it is currently unclear which additional genes are targeted by these factors during oenocyte specification, candidate factors that have been previously implicated to lie downstream of EGF signaling in oenocyte development include the Spalt-major (Salm) zinc finger protein, the Seven-up (Svp) and Hepatocyte Nuclear Factor 4 (HNF4) hormone receptors, and the ventral-veinless (Vvl) POU homeodomain protein [22,24,26].
Several fundamental questions arise from these studies such as how many oenocyte target genes are regulated by the Abd-A Hox and Pnt factors, and how are these factors integrated to regulate target gene expression during oenocyte development? Studies focused on other Hox regulated morphological structures and cell types suggest Hox factors commonly regulate many targets within a gene regulatory network [8,9,11]. For example, studies on how Ultrabithorax (Ubx) promotes haltere development and how Abdominal-B (Abd-B) specifies genitalia in Drosophila reveal that Hox factors are likely to regulate numerous target genes throughout the network [19,20,46]. Moreover, while we currently do not know how Abd-A and the Pnt transcription factors are physically integrated, a recent study using bimolecular fluorescence complementation (BiFC) found that Abd-A and Pnt can interact in Drosophila embryos [47]. Hence, future studies focused on the transcriptional mechanisms used by Abd-A and Pnt factors is likely to shed new insight into the underlying gene regulatory networks required for the development of these hepatocyte-like cells.

Plasmid and transgenic fly generation
The PntP1OE cis-regulatory element (3R: 23309267. . .23309966 R6.13) was amplified by PCR from genomic DNA, and PntP1OE mutations and deletions were introduced by PCR mediated mutagenesis. All enhancers were cloned into the pLacZ-attB plasmid and confirmed by DNA sequencing. Transgenic fly lines were generated by FC31 integration into the 51C insertion site [48] (Injections by Rainbow Transgenics).

Protein purification and EMSAs
The following proteins were purified from BL21 bacterial cells as previously described [51]: His-tagged Abd-A; His-Antp; His-Hth and untagged Exd heterodimers [52,53]. DNA probes were annealed from primers labeled with IRdye700 at 5'terminus [54]. EMSAs were performed as previously described using native polyacrylimide gel electrophoresis. The acrylamide gels were imaged using a Odyssey LiCOR cLX scanner.