qPCR primer sequences are found in Table S1. All cloning primer sequences are presented in Table S2 and siRNA sequences in Table S3.

The Glossina morsitans morsitans colony maintained in the insectary at Yale University was originally established with puparia from fly populations in Zimbabwe. Newly emerged flies are separated by sex and mated at three to four days post-eclosion. Flies are maintained at 24±1°C with 50–55% relative humidity, and receive defibrinated bovine blood every 48 h using an artificial membrane system [17].

Samples for gene expression analysis during parturition were collected as follows. Female flies pregnant with 3^rd instar larvae were collected. Flies were checked every 24 hours and individuals that had undergone parturition during the 24 hour period were collected to stage them in sample groups. Three flies were collected at the following time points, pregnant (3^rd instar larva), 0–24, 24–48, 48–72, 72–96, 96–120, 120–144, 144–168, and 168–192 hrs post parturition (pp). RNA and protein were isolated from flash frozen female flies and tissue samples utilizing the standard Trizol (Invitrogen, Carlsbad CA) protocol with a modification to the final step in which the isolated protein pellets are dissolved in cracking buffer (8M urea, 3M thiourea, 1% dithiothreitol (DTT) and 4% CHAPS) [18]. cDNA was prepared from total RNA using the Invitrogen Superscript III kit (Invitrogen). Levels of mgp1 were determined by CFX Connect Real Time PCR Detection System with SYBR Green Supermix (Bio-Rad, Hercules, CA). The data were analyzed with software version 3.1 (Bio-Rad). All samples were normalized according to Glossina tubulin (tub) expression levels.

Western blotting was performed utilizing the protein from the samples described above using previously described antisera and protocols [6], [19].

Genomic sequences and automated annotations were derived from the recently completed G. m. morsitans genome [16] and are available at Vectorbase (http://gmorsitans.vectorbase.org/Glossina_morsitans/Info/Index).

Two constructs were created using 2 kB and 0.5 kB of the 5′ upstream from the mgp1 predicted transcription start site. These fragments were PCR amplified from G. m. morsitans genomic DNA using Platinum Taq DNA Polymerase High Fidelity (Invitrogen) and sub-cloned into the T-vector cloning vector (Promega, Madison WI). Inserts were excised from the T-vector using SphI and NotI and ligated into the p-element vector pPelican [20] between the SphI and NotI sites. This vector includes a lacZ reporter gene downstream of the regulatory sequence. The enhancer/reporter constructs were used to transform w¹¹¹⁸ flies by p-element mediated transformation via a commercial transformation service (Best Gene: Drosophila Injection Services (http://www.thebestgene.com). The transformations generated 4 lines carrying the 2.0 kB construct mgp1-Bgal-2.0-1-4 and 3 lines carrying the 0.5 kB construct mgp1-Bgal-0.5-1-3. Drosophila lines were maintained according to standard protocols [21].

Tissues from male and female Drosophila were dissected in 1× PBS. Dissected tissues were fixed and stained for β-galactosidase activity with the ß-galactosidase Staining Kit (Mirus, Madison WI) following the manufacturer's protocols.

Flies expressing an eGFP reporter with 509 (mgp1-egfp-509), 236(mgp1-egfp-236), 112 (mgp1-egfp-112) or 13 bp (mgp1-egfp-13) nucleotide versions of the mgp1 upstream sequence were generated utilizing the PhiC homologous recombination transformation system [22], [23]. The 509 bp mgp1 upstream fragment was PCR amplified from Glossina morsitans genomic DNA using Platinum Taq DNA Polymerase High Fidelity (Invitrogen) and sub-cloned into the T-vector cloning vector (Promega). The fragment was excised with SphI and SpeI then ligated into the nuclear EGFP enhancer analysis vector pStinger [20] between the SphI and NheI sites by standard cloning methods. The mgp1-pStinger construct was used as a template to amplify the 509, 236, 112 and 13 base pair variants of the mgp1 promoter-egfp fusion constructs. The forward and reverse primers for the constructs were tailed with the AttB40 PhiC recombination site sequence. Amplified constructs were cloned into T-vector and sent for transformation via a commercial transformation service (Best Gene: Drosophila Injection Services (http://www.thebestgene.com). Constructs were injected into the Drosophila line genotype y¹ w^*; P[attP.w⁺.attP]JB53F (Bloomington Stock Center #27386). The AttB integration site is 53F8, 2R:12985015 [22]. Surviving injected flies were crossed with CyO/Sco 2^nd chromosome balancer flies and screened for loss of eye color as a negative marker of transgene insertion. Drosophila lines were maintained according to standard protocols [21].

RNA was isolated from flies using Trizol (Invitrogen). cDNA was prepared from total RNA using the Invitrogen Superscript III kit (Invitrogen). Transgene transcript levels were quantified by qPCR with iQ SYBR Green Supermix using egfp specific primers and the CFX Connect Real Time PCR Detection System (Bio-Rad). Data was analyzed using software version 3.1 (Bio-Rad). All treatments were normalized according to Drosophila beta-tubulin expression levels using gene specific primers and carried out in triplicate. Visual inspection of transgene expression was performed by tissue dissection in 1× PBS followed by fluorescent microscopy using a Axio Cam dissecting scope equipped with a X-cite series Q fluorescence system (Zeiss Microscopy & Image Analysis, Thornwood, New York).

Stage specific quantification of egfp transgene expression was performed as described above with total RNA from 3 independent groups of larval, pupal, and 3–5 day old male and female flies from the mgp1-egfp-509 line. 3–5 day old adult females from the mgp1-egfp-236, mgp1-egfp-112 and mgp1-egfp-13 lines were also collected and analyzed in the same manner to compare transgene expression between the lines.

Nutritional deprivation experiments were performed by maintaining the mgp1-egfp-509 line on either complete (yeast extract 100 g/L, sucrose 100 g/L, Agar 27 g/L, Methlyparaben: 30 ml of 100 g/L, Proprionic acid 3 mL/L) or minimal media (yeast extract 12.5 g/L, sucrose 12.5 g/L, Agar 27 g/L, Methlyparaben: 30 ml of 100 g/L, Proprionic acid 3 mL/L) in 25×95 mm polystyrene vials. Egg deposition was monitored daily for each group. Parallel treatments were run to quantify egfp levels in stressed flies using the qPCR methods described above.

Putative transcription factor binding sites were predicted within the mgp upstream regulatory region using the “MatInspector” program from Genomatix (http://www.genomatix.de/) [24]. Upstream sequences from other characterized and predicted milk proteins were derived from the draft Glossina genome (www.vectorbase.org) [16]. The sequences used were from the 124 bp mgp1 (GMOY009745) critical region, trf (GMOY004228), asmase1 (GMOY002246), mgp2-10 (GMOY012368, GMOY012125, GMOY001342, GMOY012016, GMOY001343, GMOY012016, GMOY012369) genes. 500 bp from the predicted transcription start site of each gene was used in a comparative analysis. This analysis was also performed using MatInspector.

Predicted Glossina homeodomain sequences were identified from the Glossina genome [16]. Protein sequences from the 104 known Drosophila homeodomain factors were used to search the Glossina genomic assembly and the de novo assembly from [12] by tBLASTn search. All resulting hits with scores lower than 1×10⁻¹⁰ were collected and searched against the NCBI database by BLASTx search. Sequences were then annotated with nomenclature from the highest scoring hit result.

We have recently completed a transcriptome study focusing on differences between pregnant/lactating or dry flies (non-lactating) [12]. Raw sequence data are available from the sequence read archive at NCBI. We utilized the RNA-seq reads from that study to compare the transcript abundance of Glossina homeodomain proteins in lactating and dry flies. RNA-seq data were mapped directly to the Glossina homeodomain genes utilizing CLC Genomics Workbench (CLC bio, Cambridge, Massachusetts). Normalized gene expression was calculated as RPKM [25]. Statistical differences of RPKM values between samples were determined by Kal's test following Bonferroni correction [26]. Statistical significance was determined at P<0.05 and a two-fold higher or lower transcript abundance in lactating flies was utilized to determine genes of interest [27].

Five candidate Glossina homeodomain genes were analyzed by qPCR based tissue and stage specific expression analysis. Tissue-specific samples were acquired from females harboring 2nd instar larva. qPCR was conducted as before and transcript levels were normalized to tubulin.

Functional analysis of the ladybird late homeodomain factor was performed using synthesized siRNAs homologous to the Glossina ladybird late ortholog. siRNAs were designed by web-based tools (IDT, Coralville, IA) and ordered commercially (IDT). The siRNA consists of two Duplex sequences for lbl. Control siRNA molecules were designed as complementary to green fluorescent protein (GFP) mRNA (IDT, Cordville, IA). Concentration of each siRNA was adjusted to 700–750 ng/µl in PBS with a Nanodrop spectrophotometer. Mated female flies were injected with 1.5 µl siRNA 6–8 d after adult emergence. siRNA was shown to have no discernible effect on larval transcripts [28]. Expression levels were determined by qPCR (as described; normalized to tubulin) 5 d after injection for lbl and 8 d after injection for mgp1.

Transcript abundance of the mgp1 gene is related to intrauterine larval development. Previous expression data for mgp1 was based upon samples staged by female reproductive physiology and larval development [15]. Based upon these data, mgp1 transcripts appear to decrease after larval deposition (parturition) and increase again at the onset of intrauterine larval development in the next pregnancy cycle. To obtain a high resolution temporal analysis of milk protein gene expression during this period, pregnant flies were collected and synchronized by time of parturition in 24 hour intervals rather than by pregnancy status. Transcript levels from these samples were quantified by qPCR analysis, and protein expression was determined by Western analysis. The results show that mgp1 transcript abundance and protein levels undergo a precipitous drop after parturition and reach their lowest levels at 24–48 hours post-parturition (Figure 1A and 1B). Transcript and protein levels begin to increase again between 48–72 hours post parturition and continue to increase to the last time point in the series at 168–192 hours post parturition when there is a 2^nd instar intrauterine larva in the uterus.

Analysis of the regulation of mgp1 expression required the development of genetic tools that facilitate the observation of Glossina promoter function in vivo. To accomplish this, we leveraged the genetic tools available in D. melanogaster to perform an in vivo analysis of the mgp1 promoter. Drosophila is related to Glossina as both are members of the dipteran “Higher Flies” (Brachycera) suborder.

The transformation constructs were developed using the pPelican enhancer/reporter plasmid [20]. The plasmid contains a β-galactosidase (β-gal) reporter gene and the promoter/reporter fusion is flanked by gypsy insulator sequences to prevent the influence of local regulatory elements. Transformation was accomplished by p-element transposition. Reporter expression was visualized by staining for β-gal activity in dissected transgenic Drosophila tissues. Two constructs were created including either 2 kB or 0.5 kB sequence of the 5′ upstream from the predicted mgp1 transcription start site (Figure S1A). Transformation of these constructs resulted in the production of four transgenic lines for the 2.0 kB construct (mgp1-β-gal-2.0-1 to -4) and three transgenic lines for the 0.5 kB construct (mgp1-β-gal-0.5-1 to -3).

In tsetse, mgp1 expression is specific to the milk gland of adult female flies [11]. We analyzed different tissues from both mgp1-β-gal-2.0 and mgp1- β-gal-0.5 lines with β-gal staining to understand the sex and tissue specific nature of the reporter gene expression. Transgene expression was exclusive to the accessory glands (paraovaria) of the female reproductive tract in both transgenic lines (Figure 2A) while this staining pattern was not observed in control Drosophila. Background staining was observed in the uterus, midgut and in specific cells along the midline of the dorsal abdominal surface in both sexes. This background staining pattern was also observed in untransformed control flies suggesting that this endogenous β-gal activity may result from populations of bacteria resident within the fly. The specific staining observed in the accessory glands of females suggests that regulatory mechanisms and factors driving sex and tissue specific expression of mgp1 in tsetse are conserved in Drosophila. It also demonstrates that the transcriptional regulatory elements required for sex and tissue specific expression are contained within the 0.5 kB upstream region of the mgp1 gene.

To identify the minimal region required for tissue and stage specific gene expression by the mgp1 promoter, additional transgenic Drosophila lines were generated. To eliminate the confounding issues of background staining with endogenous β-gal activity and variation of transgene expression due to position effect, an alternative transformation strategy was utilized. The new transgenic lines were generated using the PhiC recombination system [22], [23]. This system allows for the stable incorporation of transgenes into a specific genomic locus via homologous recombination, thereby eliminating position effect derived variation in expression between lines. To enhance transgene detection and reduce background, the β-gal reporter gene was replaced with the nuclear specific enhanced green fluorescent protein (EGFP) reporter gene from the pStinger p-element transformation vector. We generated four transgenic lines, which include either 509 (mgp1-egfp-509), 236 (mgp1-egfp-236), 112 (mgp1-egfp-112) or 13 bp (mgp1-egfp-13) segments corresponding to the mgp1 regulatory region upstream of the start site (Figure S1B).

All life stages and tissues from these lines were examined by florescent microscopy for nuclear specific EGFP expression. Microscopic analysis of EGFP florescence revealed reporter gene expression is only detectable within the accessory glands of reproductively active (undergoing oogenesis and ovulation) females from the mgp1-egfp-509 line. This is almost identical to the observations in the β-gal expressing lines (Figure 2B). Of the four lines, the mgp1-egfp-509 and mgp1-egfp-236 lines were positive for female accessory specific EGFP expression by visual screening, while EGFP expression in mgp1-egfp-112 and mgp1-egfp-13 bp constructs was undetectable by visual observation. These results suggest that the minimal region for tissue specific expression and basic promoter function lies within the 236 bp upstream of the mgp1 transcription start site. The 124 bp differential region between the mgp1-egfp-236 and mgp1-egfp-112 lines appears to carry a transcriptional response element critical to the function of this promoter as a whole.

Quantitative analysis of egfp expression during different developmental and sexual stages in mgp1-egfp-509 reveals that the transgene is only significantly expressed within sexually mature adult females with only background levels present in larvae and males (Figure 2C). Comparison of egfp expression between the four lines confirms the results we obtained by microscopy analysis. Expression of the transgene is highest in mgp1-egfp-509 (Figure 2D). Transgene expression in mgp1-egfp-236 shows that the overall transcript level is significantly lower than that observed in mgp1-egfp-509. However, mgp1-egfp-236 maintains the sex and tissue specific characteristics of the mgp1-egfp-509. Transgene expression in mgp1-egfp-13 and mgp1-egfp-112 was equivalent to levels observed in male flies suggesting that regulatory function is compromised in these lines.

The reproductive cycle in Drosophila differs from that of tsetse in that oogenesis and embryo deposition occur at a steady rate once the female is sexually mature and mated as opposed to the defined cycles of oogenesis, embryogenesis and larvigenesis observed in tsetse. Nutritional stress in Drosophila results in reduction or termination of oogenesis. In the absence of appropriate nutritional stimulation yolk protein gene expression and oocyte development cease. In some cases developing oocytes may undergo apoptosis and reabsorption [29], [30]. To determine if a nutritionally induced reduction in oogenesis and ovulation would influence transgene expression, 3 day old mated female mgp1-egfp-509 flies were put on a minimal media diet followed by daily observation of egg deposition and egfp transcript levels. Maintenance on minimal media relative to complete media resulted in a significant reduction in egg production in the mgp1-egfp-509 flies within 5 days (Figure 2E). Quantitative analysis of reporter egfp expression in flies maintained on minimal media during the 5 day time period shows a direct correlation between transgene expression and egg deposition over time (Figure 2F). These results correlate the down regulation of gene expression activity within the accessory gland with reduced nutritional status/oogenesis. Whether the gland is responding to reduced nutritional stimuli or reduced reproductive stimuli is not yet understood.

We next set out to predict transcription factor binding sites that may lie within the 0.5 kb promoter region of mgp1 responsible for the sex and tissue specific expression profile we observed in both taxa. Analysis of the 0.5 kB regulatory sequence using transcription factor binding site prediction software resulted in a total of 41 predicted binding sites from a variety of factors. The matrices used in the identification included those for known insect transcription factors and eukaryotic basal promoter elements. To reduce the number of candidate factors, we narrowed our analysis to the 124 bp region between lines, mgp1-egfp-236 and mgp1-egfp-112, which is required for sex and tissue specific function. This analysis reduced the number of predictions to 11 candidate binding sites. The sites predicted include generic Drosophila homeodomain binding sites, including (DHOM: 2 sites), OVO transcription factor (DOVO: 1 site), Iroquois factor group (IRXF: 1 site), Tailless (DTLL: 1 site), Abdominal B (ABDB: 1 site), Paired homeodomain factors (PRDH: 2 sites), Dead Ringer (DRIF: 1 site), Giant (DGTF: 1 site) and Heat shock factors (DHSF: 1 site).

Previous characterizations, annotation of the Glossina genome [16] and RNA-Seq based analysis of lactating and non-lactating flies has identified 12 milk protein genes that function in the lactation process in tsetse, including mgp1, mgp2-10, trf and asmase1 [12]. Comparative analysis of the 124 bp mgp1 region critical for transgene expression in Drosophila with the 500 bp sequence upstream of the transcription start site of these genes predicts that only homeodomain protein binding sites (DHOM) are common between all 12 sequences (p-value: 0.0241) (Figure 3). The only other site in common between these promoters is that of the TATA binding protein (TATA box).

To identify homeodomain type regulatory factors that may recognize the conserved DHOM sites, we annotated and collated all sequences in Glossina bearing homology to the 106 annotated Drosophila homeodomain proteins. This was accomplished by tBLASTn search of the Glossina genome assembly [16] and a de novo assembly of RNA-seq libraries with Drosophila homeodomain protein sequences. All significant hits were submitted to a BLASTx comparison against the NCBI protein database. This analysis predicted a total of 96 putative tsetse homeodomain protein orthologs within the current genome assembly (see Table S36 in the Glossina genome paper).

Functional analysis of all candidate homeodomain factors was prohibitive. To reduce the number of targets, the expression profile of the 96 homeodomain genes were screened by RNA-Seq analysis using the available Illumina sequencing data from lactating and non-lactating flies [12]. To identify differential transcript abundance between the two physiological states, we selected genes showing a relative transcript difference of greater than 2 fold higher or lower between samples, and sequence representation by at least 500 reads between libraries. Based upon these criteria, our analysis identified 5 homeodomain genes; 3 with high transcript abundance nubbin (nub – VectorBase accession: GMOY012175), teashirt (tsh – VectorBase accession: GMOY011052) and vismay (vis – VectorBase accession: GMOY000857) and 2 with lower transcript abundance ladybird late (lbl – VectorBase accession: GMOY004068), and pox meso (poxm – VectorBase accession: GMOY002525) (Table 1).

Given the tissue specific nature of expression for mgp1 (and the other milk proteins), homeodomain factors specific to that tissue were of primary interest. We performed a qPCR based expression analysis of the five candidate factors using RNA from different tissue samples obtained from lactating female flies. The expression patterns were unique for each gene. Of the genes analyzed, only the Ladybird Late gene (lbl) displayed a milk gland/fat body specific expression pattern. The fat body and milk gland are grouped together due to their interconnected nature which makes separating them by dissection next to impossible. Transcripts for lbl were ∼30 fold higher in the milk gland/fat body tissue than in any other tissue (Figure 4A). The other four factors were found in multiple tissues at different levels of transcript abundance (Figure S2).

To determine lbl involvement in milk protein gene regulation, siRNA was utilized to perform gene knockdown analyses. Flies were treated with siRNAs against lbl, or gfp (control). The siRNA injections resulted in significant reductions in target gene transcript abundance. The levels of lbl were reduced to ∼30% of controls (Figure 4B). Following siRNA treatment, transcript abundance of the mgp1, mgp2 and asmase1 genes was measured by qPCR. The knockdown of lbl, resulted in a significant decrease of mgp1, mgp2 and asmase transcript levels relative to the GFP controls (Figure 4C). These data indicate that the lbl homeodomain protein functions as a positive regulator of the mgp1 gene either directly or indirectly.

We monitored the larviposition of siLbL treated mated females over the first gonotrophic cycle. The lbl knockdown flies showed a significant reduction in the average rate of larviposition per fly per day relative to controls (Figure 5A). Cumulative larval deposition rates between the control and lbl knockdown group revealed that the lbl group birthed about half as many larvae as the controls (Figure 5B). The reduction in fecundity in these flies is likely due to the reduced level of milk protein gene expression resulting from the knockdown of the lbl. This results in nutritional deprivation of developing larvae, larval death and abortion. This knockdown does not result in a complete disruption of tsetse fecundity in terms of the number of larvae developed per female; however we believe that this is due to incomplete penetrance of our injectable siRNA system in Glossina.