Analysing bioelectrical phenomena in the Drosophila ovary with genetic tools: tissue-specific expression of sensors for membrane potential and intracellular pH, and RNAi-knockdown of mechanisms involved in ion exchange

Background Changes in transcellular bioelectrical patterns are known to play important roles during developmental and regenerative processes. The Drosophila follicular epithelium has proven to be an appropriate model system for studying the mechanisms by which bioelectrical signals emerge and act. Fluorescent indicator dyes in combination with various inhibitors of ion-transport mechanisms have been used to investigate the generation of membrane potentials (Vmem) and intracellular pH (pHi). Both parameters as well as their anteroposterior and dorsoventral gradients were affected by the inhibitors which, in addition, led to alterations of microfilament and microtubule patterns equivalent to those observed during follicle-cell differentiation. Results We expressed two genetically-encoded fluorescent sensors for Vmem and pHi, ArcLight and pHluorin-Moesin, in the follicular epithelium of Drosophila. By means of the respective inhibitors, we obtained comparable effects on Vmem and/or pHi as previously described for Vmem- and pHi-sensitive fluorescent dyes. In a RNAi-knockdown screen, five genes of ion-transport mechanisms and gap-junction subunits were identified exerting influence on ovary development and/or oogenesis. Loss of ovaries or small ovaries were the results of soma knockdowns of the innexins inx1 and inx3, and of the DEG/ENaC family member ripped pocket (rpk). Germline knockdown of rpk also resulted in smaller ovaries. Soma knockdown of the V-ATPase-subunit vha55 caused size-reduced ovaries with degenerating follicles from stage 10A onward. In addition, soma knockdown of the open rectifier K+channel 1 (ork1) resulted in a characteristic round-egg phenotype with altered microfilament and microtubule organisation in the follicular epithelium. Conclusions The genetic tool box of Drosophila provides means for a refined and extended analysis of bioelectrical phenomena. Tissue-specifically expressed Vmem- and pHi-sensors exhibit some practical advantages compared to fluorescent indicator dyes. Their use confirms that the ion-transport mechanisms targeted by inhibitors play important roles in the generation of bioelectrical signals. Moreover, modulation of bioelectrical signals via RNAi-knockdown of genes coding for ion-transport mechanisms and gap-junction subunits exerts influence on crucial processes during ovary development and results in cytoskeletal changes and altered follicle shape. Thus, further evidence amounts for bioelectrical regulation of developmental processes via the control of both signalling pathways and cytoskeletal organisation.


(Continued from previous page)
Conclusions: The genetic tool box of Drosophila provides means for a refined and extended analysis of bioelectrical phenomena. Tissue-specifically expressed V mem -and pH i -sensors exhibit some practical advantages compared to fluorescent indicator dyes. Their use confirms that the ion-transport mechanisms targeted by inhibitors play important roles in the generation of bioelectrical signals. Moreover, modulation of bioelectrical signals via RNAi-knockdown of genes coding for ion-transport mechanisms and gap-junction subunits exerts influence on crucial processes during ovary development and results in cytoskeletal changes and altered follicle shape. Thus, further evidence amounts for bioelectrical regulation of developmental processes via the control of both signalling pathways and cytoskeletal organisation.
In Drosophila, at least two distinct pathways are responsible for planar cell polarity. One pathway depends on Dsh/Fz and acts in the wing and eye [13], while the second pathway depends on interaction of the cytoskeleton with the extracellular matrix in ovarian follicle cells (FC) [18]. FC display microfilaments at their basal side (bMF) that are oriented perpendicular to the anteroposterior (a-p) axis of the developing follicle [5,[17][18][19]. Proper bMF-orientation requires integrins as well as planar-polarised distribution of the receptor-tyrosine phosphatase Lar. Lar is known to be involved in signalling between the extracellular matrix and the actin cytoskeleton [20,21]. The first mutant shown to disrupt polarisation of bMF in FC was called kugelei, due to its prominent round-egg phenotype [22].
While bioelectrical phenomena, like gradients of V mem and pH i , become increasingly accepted as regulators of development, the mechanisms by which these signals exert influence on developmental pathways are poorly understood. Therefore, it is necessary to identify the iontransport mechanisms involved in generation and modification of the bioelectrical signals. During Drosophila oogenesis, the exchange of protons, potassium ions and sodium ions is primarily responsible for stage-specific V mem -and pH i -patterns as well as for extracellular currents [23][24][25][26][27][28]. Moreover, in the planar cell-polarity pathway of the Drosophila wing and eye, a need for bioelectrical cues to conduct signalling has been demonstrated [13,29].
The DEG/ENaC-family represents one of the largest ion-channel families in Drosophila [30]. In vertebrates, amiloride-sensitive Na + -channels have been implicated in some early developmental events, like blocking secondary sperm entry in Xenopus eggs or generating the blastocoel [31]. Members of the DEG/ENaC-family mediate Na + -absorption across the apical membrane of epithelia; they are essential for Na + -homeostasis, and are expressed in gonads and neurons [32][33][34].
In insects, proton-pumping V-ATPases are located in apical membranes of almost all epithelial tissues, where they energise secondary active transport processes [35,36]. Moreover, they are responsible for the acidification of cytoplasmic vesicles, e. g., in the follicular epithelium (FE) of Drosophila [3,16,27]. In Drosophila ovarian follicles, an involvement of V-ATPases in bioelectrical phenomena has been supposed [27,37]. In particular, the asymmetrical accumulation of V-ATPases on one side of the follicle points to a role in regulating spatial coordinates [3,37]. Several studies demonstrated that V-ATPases are also required for Notch and wingless signalling in Drosophila [29,38,39].
In Drosophila follicles, germline and soma cells are interconnected via gap junctions [40]. Members of the innexin family are known to represent the main gapjunction proteins in invertebrates [41,42]. In the Drosophila ovary, innexins 1 to 4 have been shown to be involved in the formation of different types of gap junctions [43,44]. Gap junctions can propagate alterations of V mem and pH i between germline and soma cells [3,40,44].
In the present study, we used, for the first time, genetically-encoded sensors for V mem and pH i in combination with specific inhibitors of ion-transport mechanisms in order to refine and extend earlier studies using electrophysiological recordings [23,24] or V mem -and pH i -sensitive fluorescent dyes [5,16] in the ovary of Drosophila. Out of a large number of available genetically-encoded V mem -indicators (GEVIs) with the voltage-sensing domain (VSD) of Ciona intestinalis, we chose a member of the ArcLight family. GEVIs of this family display a relatively high sensitivity as well as slow kinetics of activation and inactivation [45,46]. These characteristics appeared to be useful for analysing slow V mem -changes as in the FE of Drosophila. The selected pH i -sensor comprises a fusion of pHluorin and the Moesin actin-binding domain; it was initially designed for the visualisation of apoptotic cell-phagocytosis [47]. Due to tissue-specific expression, genetically-encoded sensors provide some advantages compared to other methods used to identify ion-transport mechanisms involved in V mem -and pH i -regulation [48,49]. In order to refine and extend the knowledge obtained using inhibitors of ion exchange, we performed a RNAi-knockdown screen of genes coding for ion-transport mechanisms and gapjunction subunits that, via V mem -and pH i -changes in the FE, might have impact on the development of the ovary and/or on oogenesis in Drosophila.

Results
Genetically-encoded sensors of V mem and pH i reliably respond to inhibitors of ion-transport mechanisms V mem -sensor ArcLight and pH i -sensor pHluorin-Moesin Two genetically-encoded fluorescent V mem -and pH isensors, ArcLight and pHluorin-Moesin, in combination with six inhibitors (cf. [5,16]) were used to analyse the roles that specific ion-transport mechanisms play in regulating V mem and pH i in the follicular epithelium of Drosophila during stage S10B.
ArcLight-family GEVIs respond to depolarisation upon blue-light excitation with reduced green fluorescence of superecliptic pHluorin, while they respond to hyperpolarisation with enhanced green fluorescence (Fig. 1e). Superecliptic pHluorin is protonated at relatively depolarised V mem (dark or "ecliptic") and mostly deprotonated at relatively hyperpolarised V mem (bright). The pH i -sensor pHluorin-Moesin emits green light upon blue-light excitation as well. Due to protonation, it responds to relative acidification with reduced fluorescence, whereas, due to deprotonation, relative alkalisation is indicated by enhanced fluorescence (Fig. 1e).
The specific expression of both ArcLight and pHluorin-Moesin at the FC cortex revealed, during the course of vitellogenesis (S8-S12), stage-specific patterns of V mem and pH i (Fig. 1b and c) which are comparable to those obtained previously with the fluorescent indicator dyes DiBAC 4 (3) and 5-CFDA,AM (cf. [3,5,16,17]). Uniform FE-specific expression of the sensors was controlled by the tj-Gal4-driven expression of GFP (Fig. 1a).
In the pHluorin-expressing FE, the inhibitors concanamycin A, 9-anthroic acid, furosemide and glibenclamide, respectively, led to higher fluorescence intensities (alkalisation), whereas amiloride led to lower fluorescence intensity (acidification). While the strongest effect on pH i was observed with 9-anthroic acid, verapamil showed no significant effect (Fig. 2b).
To directly compare the effects of all inhibitors on either V mem or pH i , a mean intensity ratio of the experimental and the control groups of four experiments was calculated for each treatment (Figs. 3 and 4). This evaluation considered the variability between experiments with the same treatment, whereas the evaluation shown in Fig. 2 considered the variability between different follicles. Both evaluations disclosed inhibitory effects with the same tendency on V mem and pH i . In addition, they confirmed the results of previous studies [3,5,16] showing that the targeted ion-transport mechanisms are involved in the regulation of bioelectrical properties in the FE of Drosophila.

RNAi-knockdowns of ion-transport mechanisms and gapjunction subunits affect ovary development and oogenesis
The purpose of our screen was to investigate whether RNAi-knockdowns of candidate genes of ion-transport mechanisms or gap-junction subunits result in longterm effects on ovary development and/or on oogenesis. In particular, we wanted to see if RNAi-knockdowns exert influence on the FE-specific cytoskeleton in a similar way as various inhibitors of ion-transport mechanisms [16]. For RNAi-knockdown in the FE, we combined VDRC UAS-strains or TRIP UAS-lhRNAand UAS-shRNA-strains of relevant genes with the soma-specific tj-Gal4 driver line. In addition, we used the germline-specific mat-tub-Gal4 or MTD-Gal4 driver lines for RNAi-knockdown in NC and Ooc (see Fig. 5, Table 1 and Additional file: Table S1). As controls, ovaries from flies expressing the UAS-constructs at low levels in the germline were used (e. g., mat-tub-Gal4-GeneSwitch > ork1 shRNA). As expected, these ovaries did not show any phenotype differing from wt (Fig. 6d).
We identified five genes of ion-transport mechanisms and gap-junction subunits showing effects on ovary development and/or on oogenesis (Table 1, Figs. 5 and 6): RNAi of vha55 (subunit B of V-ATPase) caused, via soma knockdown, size-reduced ovaries with degenerating follicles from S10A onward or, via germline knockdown (depending on the RNAi-construct), NC rests in S11-S14 and excess of FC around the micropyle in S14. Complete loss of ovaries or size-reduced ovaries (some follicles but no ovarioles discernible) were the results of soma knockdowns of inx1 or inx3 (innexin; gap-junction subunit). In addition, germline knockdown of inx1 (via mat-tub-Gal4) led to degenerating follicles of all vitellogenic stages.

RNAi-knockdowns of the genes rpk and ork1 show striking effects
The strongest RNAi-knockdown effects were observed for rpk (ripped pocket), a member of the DEG/ENaC (epithelial sodium-channel) family: Reduced transcript levels of Fig. 1 Analysis of V mem and pH i using the genetically-encoded fluorescent V mem -and pH i -sensors ArcLight and pHluorin-Moesin, respectively. a Uniform expression of GFP in the FE using the soma-driver tj-Gal4 (control); WFM-images of typical tj-Gal4 > UAST-GFP follicles of vitellogenic stages S8-S12 (scale bars represent 50 μm). b, c Pseudocolour images of follicles expressing ArcLight (b) or pHluorin-Moesin (c) in the FE; median optical sections (SIM) of typical follicles of S8-S12 (scale bars represent 50 μm). d Analysis of fluorescence intensities in the FE ("mean grey value"; area marked in yellow); examples of four follicles of S10B (and one of S11) expressing pHluorin-Moesin (WFM-image; scale bar represents 200 μm). e Both the V mem -sensor ArcLight and the pH i -sensor pHluorin-Moesin use the chromophore pHluorin which responds, in deprotonated state, to blue-light excitation (blue arrow) with the emission of green light (green arrow). The exact mechanism of ArcLight is not known, but is believed to involve voltage-dependent dimerisation leading to protonation of the chromophore (VSD, voltage-sensing domain; scheme inspired by [45]). In cells expressing the respective sensor, relative depolarisation or relative acidification is indicated by weaker fluorescence intensities, whereas relative hyperpolarisation or relative alkalisation is indicated by stronger fluorescence intensities rpk in the FE resulted in complete loss of ovaries, whereas reduced levels in the germline (via MTD-Gal4) led to sizereduced paired or single ovaries showing discernible ovarioles, but no follicles. Via the mat-tub-Gal4 driver, follicles of S11-S14 with NC rests, and follicles of S14 with excess of FC around micropyle were obtained. Females with RNAi-knockdown of ork1 (open-rectifier K + channel 1) in the soma produced spherical follicles, resembling the kugelei mutant ( Fig. 6a-c, cf. [22]). This phenotype was especially prominent in follicles older than S10B. Compared to wt, ork1-follicles revealed alterations in the organisation of the bMF-and MT-patterns in the FE (Fig. 6e-f).
As described previously [5,17,19], the bMF of wtfollicles are polarised perpendicular to the a-p axis (transversal alignment), especially during S8, S10A and S12. On the other hand, the MT of wt-follicles are characterised by a-p alignment in centripetal FC (cFC) in S9, as well as in cFC and mainbody FC (mFC) in S10B.
In ork1 follicles, however, no a-p alignment of MT was detected in any analysed stage, while the overall MTpattern is less dense and less polarised compared to wt (Fig.  6e). On the other hand, typical condensations of bMF ( Fig.  6f), as in wt cFC in S9, are missing in ork1. Although bMFbundles in ork1 S10A show the same parallel transversal alignment as in wt, the overall bMF-cytoskeleton appears to be weaker in some areas. In contrast to wt, ork1 S10B and S12 are characterised by disturbed transversal bMFalignment, showing parallel bundles within FC, but chaotic organisation relative to neighbouring FC. The degrees of cytoskeletal alterations vary between different ork1-follicles of the same stage and between different areas in the same follicle. Taken together, during vitellogenic stages, wtfollicles show characteristic longitudinal MT-and transversal bMF-alignments and an elongated follicle shape (cf. [5,17]). In contrast, ork1-follicles are characterised by disturbed MT-and bMF-alignments and a spherical follicle shape, resembling the cytoskeletal organisation and follicle shape in round-egg mutants [18,22,50].

V mem -and pH i -changes in the FE revealed by tissuespecifically expressed sensors
We have shown that the genetically-encoded sensors ArcLight and pHluorin-Moesin respond to bioelectrical changes occurring in the FE during the course of oogenesis. Moreover, in the FE of S10B, both sensors revealed changes of V mem or pH i resulting from the inhibition of several ion-transport mechanisms that have been characterised in previous studies using various methods [3, 16, 25-27, 34, 51, 52]. Thus, our study shows that genetically-encoded sensors are reliable tools for investigations of this kind. In addition, the results lend further support to the notion that NHEs, Na + -channels, V-ATPases, ATP-sensitive K + -channels, voltage-dependent L-type Ca 2+ -channels, Cl − -channels, and Na + /K + /2Cl −cotransporters play important roles in modifying V mem and pH i in the FE of Drosophila.
While the strongest effect on V mem was observed using furosemide (Na + /K + /2Cl − -cotransporters), the weakest was observed using concanamycin A (V-ATPases). The strongest effect on pH i was obtained with 9-anthroic acid (Cl − -channels), whereas verapamil (L-type Ca 2+channels) showed no significant effect. Relatively small impact of inhibitors, as observed e. g. for concanamycin A or verapamil, is supposed to be due to compensatory effects exerted by other ion-transport mechanisms. Especially members of the V-ATPase-and DEG/ENaC-families [30,33,36] can substitute for other family members as well as for other types of ion-transport mechanisms.
Using the genetically-encoded sensors, we detected similar inhibitory effects on V mem and pH i in the FE as described previously using the voltage-and pH-sensitive fluorescent dyes DiBAC 4 (3) and 5-CFDA,AM, respectively [16,48,49]. According to both methods, the treatment with glibenclamide, furosemide or 9-anthroic acid resulted in alkalisation. Glibenclamide (ATP-sensitive K + -channels) is supposed to block H + -transport indirectly [16,35], while furosemide and 9-anthroic acid are expected to influence pH i via Cl − /HCO 3 − -antiport [6,16,53,54]. For concanamycin A, inhibiting V-ATPases [55], we observed alkalising effects in the FE. For bafilomycin A1, another inhibitor of V-ATPases, alkalisation of cytoplasmic vesicles and acidification of the cytoplasm was reported [16]. Therefore, the alkalisation observed for concanamycin A is supposed to refer to cytoplasmic vesicles not discernible with pHluorin-Moesin using WFM.
Instead of hyperpolarisation, as reported by [16], we observed strong depolarisation after treatment with 9-(See figure on previous page.) Fig. 2 Genetically-encoded V mem -and pH i -sensors reveal changes of bioelectrical properties resulting from the inhibition of ion-transport mechasms. a Schematic overview of the analysed ion-transport mechanisms and their specific inhibitors (in brackets). b All inhibitors led to significant changes of V mem and/or pH i in the FE. While inhibition of NHEs and Na + -channels or L-type Ca 2+ -channels resulted in relative hyperpolarisation, inhibition of V-ATPases, Cl − -channels, Na + /K +− /2Cl − -cotransporters or ATP-sensitive K + -channels resulted in relative depolarisation. Concerning pH i , inhibition of V-ATPases, Cl − -channels, Na + /K + /2Cl − -cotransporters or ATP-sensitive K + -channels resulted in relative alkalisation, whereas inhibition of NHEs and Na + -channels caused relative acidification. The inhibition of L-type Ca 2+ -channels had no significant effect on pH i . Normalised values of 15 < n < 21 S10B-follicles were averaged (relative intensity). Mean values, shown with their standard deviation, were compared using an unpaired t-test (* p < 0.05; *** p < 0.001) anthroic acid, furosemide, glibenclamide or concanamycin A. In the case of DiBAC 4 (3), reduced fluorescence intensity, indicating hyperpolarisation, might also be due to quenching [49], since depolarising effects of glibenclamide or bafilomycin A1 have been described [56,57]. On the other hand, in our experiments, higher inhibitor concentrations (up to × 100, compared to [16]) were necessary to reliably detect V mem -and pH i -changes with the membrane-bound genetically-encoded sensors. The observed depolarisation might, therefore, be attributed to high inhibitor concentrations representing a challenge for the cell. Correspondingly, blockers of oxidative phosphorylation and, thus, of almost all energy-dependent ion transport, like sodium azide or dinitrophenol (cf. [25]), had also depolarising effects on the FE (unpublished results). Mean values, shown with their standard deviation, were compared using a one-sample t-test (* p < 0.05; ** p < 0.01). The strongest effects on V mem were obtained with 9-anthroic acid, furosemide and glibenclamide, respectively Compared to fluorescent indicator dyes, one disadvantage of membrane-bound genetically-encoded sensors is their lower sensitivity, making longer exposure times and higher inhibitor concentrations necessary. Apart from that, these sensors provide several advantages: In combination with the Gal4-UAS-system, they allow the visualisation of V mem -or pH i -changes in the cell type of choice without any influences from adjacent cell types.
In addition, due to stable expression and low sensitivity to photo-bleaching, long-term imaging studies are more practicable. Finally, since unintentional interactions with other substances, as possible for fluorescent dyes, are reduced, shorter experimental protocols can be applied [45,58,59].
In conclusion, the use of genetically-encoded sensors and fluorescent indicator dyes [16] both revealed . Relative alkalisation is indicated by stronger (bright/white), relative acidification by weaker (dark/blue) fluorescence intensities. The experiments were repeated at least four times. b While amiloride caused a slight decrease in fluorescence intensity (acidification), concanamycin A, 9-anthroic acid, furosemide and glibenclamide led to significantly increasing fluorescence intensities (alkalisation). Verapamil showed no effect on pH i . To consider the variability between experiments, mean intensity ratios of the experimental and control groups (inhibited/control) of n = 4 experiments for each inhibitor were calculated. Mean values, shown with their standard deviation, were compared using a one-sample t-test (* p < 0.05; ** p < 0.01). The strongest effects on pH i were obtained with concanamycin A, 9-anthroic acid, furosemide and glibenclamide, respectively alterations of V mem and/or pH i in the FE. Therefore, both methods provide evidence that the targeted iontransport mechanisms play important roles in generating bioelectrical signals during oogenesis of Drosophila.
RNAi-knockdowns of a DEG/ENaC-subunit, a V-ATPasesubunit, or gap-junction subunits exert long-term effects on ovary development and/or oogenesis Due to results from inhibitor studies, it was tempting to investigate whether RNAi-knockdowns of candidate genes of ion-transport mechanisms or gap-junction subunits, showing enriched ovary expression, affect the course of ovary development or oogenesis. We found highly penetrant phenotypes for the genes rpk, vha55, inx1 and inx3. Most severe effects were obtained after RNAi-knockdown in somatic cells, indicating that the respective proteins are particularly relevant in FC. Several ion-transport mechanisms have already been related to pH i -regulation in the Drosophila ovary. It has been reported that the Na + /H + -exchanger Nhe2 is responsible for an increase in pH i during prefollicular cell differentiation [6]. In addition, ae2, a Cl − /HCO 3 − -exchanger, was identified as a regulator of pH i in the FC lineage: Loss of ae2 resulted in reduced fertility, fewer ovarioles, reduced follicle number and reduced ovary size, suggesting that this phenotype is caused by dysregulation of pH i [6]. Data from our RNAi-screen indicate an impairment of ovary development and/or oogenesis after knockdown of the DEG/ENaC-subunit RPK, the V-ATPase-subunit Vha55, and the gap-junction subunits Inx1 and Inx3.

DEG/ENaC-subunit RPK
The strongest effects after both soma-and germlineknockdown were observed for rpk. It has been reported that rpk is specifically expressed in gonads and in the early embryo, having a proposed function in gametogenesis [31,34,60]. Consequently, soma-knockdown of rpk resulted in complete loss of ovaries, whereas germline-knockdown led to size-reduced paired or single ovaries with beginning (See figure on previous page.) Fig. 5 Summary of genes coding for ion-transport mechanisms and gap-junction subunits showing effects in RNAi-knockdown. a Scheme of a S10Bfollicle: somatic cells (turquoise), germline cells (beige). For soma knockdown of relevant genes of ion-transport mechanisms and gap-junction subunits, the tj-Gal4 driver was used, whereas for germline knockdown, the mat-tub-Gal4 and MTD-Gal4 drivers were used. b Out of all performed RNAi-knockdowns (left column: soma knockdown via tj-Gal4; middle and right columns: germline knockdown via mat-tub-Gal4 and MTD-Gal4, respectively), the knockdowns of rpk (first line), inx1 (second line), inx3 (third line; left: size-reduced ovaries with single follicles, middle and right: no effects), and vha55 (fourth line; right: no effects) resulted in striking effects on ovary morphology (DAPI) and/or oogenesis (scale bars represent 100 μm). While soma knockdown of vha55 led to degenerating follicles from S10A onward, germline knockdown of inx1, vha55 and rpk resulted in NC rests and excess of FC around micropyle. The strongest effects were observed for rpk: Reduced rpk-transcript levels in the FE resulted in loss of ovaries, whereas reduced levels in the germline led to size-reduced paired or single ovaries showing ovarioles, but no follicles. c-e Germline driver-directed expression of myrGFP in plasma membranes of NC and Ooc (c: mat-tub-Gal4-GeneSwitch, d mat-tub-Gal4, e MTD-Gal4). myrGFP has an N-terminal myristoylation sequence directing GFP to plasma membranes under UASp-control. For soma driver-directed expression of GFP, see  ovariole formation, but no developing follicles. The severe phenotypes following rpk-knockdown are likely to be related with functions during larval development.
Many genes of the pickpocket family, like rpk, exhibit changing expression patterns throughout early development as well as in adult females, providing further hints Fig. 6 RNAi-knockdown of ork1 in the FE results in spherical follicles with altered cytoskeletal organisation. a-d Brightfield-images of follicles from tj-Gal4 > ork1 shRNA (a-c) and mat-tub-Gal4-GeneSwitch > ork1 shRNA ovaries (control; d). a For soma knockdown of ork1, the tj-Gal4 driver was used. Ovaries of all analysed flies contained spherical follicles and eggs (S8-S14; scale bar represents 100 μm). b and c Brightfield-images of S10Band S12-tj-Gal4 > UAS-ork1 shRNA follicles; the oocyte nucleus (dorsal) is marked with an asterisk (scale bars represent 50 μm). d Ovaries from mattub-Gal4-GeneSwitch > UAS-shRNA flies, having a low transcription level of ork1-shRNA, were used as control (scale bar represents 100 μm). Similar to ovaries from strong germline knockdowns via the mat-tub-Gal4 and MTD-Gal4 drivers (cf. Table 1), ovaries from control flies only produced follicles resembling the wt. e In contrast to wt, ork1-follicles (S9 and S10B) exhibit a weaker microtubule (MT) cytoskeleton in the FE, and the MT are not aligned along the a-p axis (scale bars represent 10 μm). Tangential optical sections (SIM) of typical anti-acetylated α-tubulin-treated follicles are shown. f Concerning basal microfilaments (bMF) in the FE, ork1 exhibits even stronger anomalies in comparison to wt. Tangential optical sections of typical S9, S10A, S10B and S12 wt-and ork1-follicles stained with fluorescent phalloidin are shown (cFC, centripetal FC; mFC, mainbody FC; vFC, ventral FC; a, anterior; p, posterior; d, dorsal; v, ventral); d-v orientation, as indicated, applies to all images, except vFC-images (scale bars represent 10 μm). Since, due to bMF-condensations in wt-follicles, it is difficult to reveal transversal bMF-alignment in dorsal and lateral cFC [5,17], vFC are shown for S10B. Typical bMF-condensations (asterisks), as in S9 cFC in wt, are missing in ork1. The bMF in ork1 S10A show the same parallel alignment as in wt, however, in some areas (arrowheads), the bMF-cytoskeleton is weaker. During S10B and S12, ork1-follicles are characterised by a disturbed transversal transcellular bMF-alignment, resembling the kugelei mutant [22] for their role in developmental signalling and morphogenesis [6,30]. Microarray-expression data from the FlyAtlas database indicate highest expression levels for rpk in ovary and testis [30]. However, rpk was not detected in ovarian stem cells and early cysts [34]. It has been suggested that rpk and related genes play a role in fluid distribution and cell-volume regulation during gametogenesis and early development [33]. Defects in volume regulation of NC and FC would explain the occurrence of NC rests and the excess of FC in S14.

V-ATPase-subunit Vha55
Soma-knockdown of vha55, coding for subunit B of the vacuolar H + -ATPase, caused size-reduced ovaries with degenerating follicles from stage S10A onward. It has been reported that genetic knockout of vha55 leads to a larval lethal phenotype [61]. V-ATPases are highly expressed in ovaries [36,62] where they are predominantly located in apical FC membranes and in the oolemma [27,37]. Moreover, V-ATPases are presumed to be involved in bioelectrical phenomena during oogenesis [3,27] as well as in osmoregulation and follicle growth by water uptake, especially during S10-S12 [27]. Due to the loss of Vha55-function, follicle growth might be inhibited in S10 and, as a consequence, degeneration might take place. On the other hand, organelleassociated V-ATPases are necessary for the acidification of cytoplasmic vesicles (cf. [5,27]). Consequently, cells lacking V-ATPase-function show impaired acidification of the endosomal compartment and fail to degrade endocytic cargoes [38]. This observation could also explain the degeneration during S10, since cargo sorting is essential for epithelial polarisation, vitellogenesis and other developmental processes [63].

Gap-junction subunits Innexin 1 and Innexin 3
For inx1, a function in somatic stem-cell formation is likely since no ovaries were found after somaknockdown. Moreover, inx1 has been shown to be predominantly expressed in FC [44]. For mutants of another gap-junction gene, inx4, it has been reported that sizereduced gonads correlate with reduced survival of differentiating early germline cells [42,64]. Our inx3-soma knockdown resulted in size-reduced ovaries, in which few follicles were discernible. Therefore, we assume influences of inx3 on FC differentiation. Rudimentary ovaries combined with impaired follicle maturation, as observed for inx3, are also found in several mutants, e. g., the transcription-factor mutant stonewall [65]. After germline-knockdown of inx3, no defects were observed, which corresponds to the predominant expression of inx3 in FC [44]. Considering that bioelectrical signals can pass, via gap junctions, from somatic cells to germline cells and vice versa [40], changes of V mem and pH i , resulting from RNAi-knockdown either in the soma or the germline, might also become transmitted to the connected tissue and exert indirect influence on development.
Our RNAi-knockdowns of inx2 and ductin had no effects on ovary or follicle morphology (see Additional file: Table S1). Previously, inx2 has been associated with defects in gametogenesis, and ductin, subunit c of V-ATPase, was expected to contribute to developmentally important bioelectrical signals [37,44,[66][67][68]. Such missing effects of RNAi-knockdown might depend on the respective RNAi-strain used since, e. g., not all tested rpk-or ork1-strains caused knockdown-effects (Additional file: Table S1). Similarly, it has been reported that loss of stim-transcripts caused severe wing defects and resulted in size-reduced wings [7]. However, in our screen, no effects of stim-knockdown could be detected in the ovary (Additional file: Table S1).
RNAi-knockdown of the potassium channel Ork1 results in spherical follicles with altered cytoskeletal organisation in the FE As a knockdown-candidate for K + -channels, we chose the gene ork1 (open rectifier K + channel 1). According to the FlyAtlas database of gene expression [52], ork1 RNA is enriched in the ovary. While soma-knockdown of ork1 resulted in altered follicle shape, germline-knockdown had no effect.
Analysis of the bMF-organisation in the FE revealed cytoskeletal peculiarities in ork1-follicles compared to wt. Wt-follicles show transversal bMF-alignment in S8-S12 (cf. [5]) and an elongated shape, whereas ork1-follicles show disturbed bMF-alignment and a spherical shape. Similar to the round-egg mutants fat2/kugelei, trc, fry, msn and Lar [18, 20-22, 50, 69], the failure to globally organise bMF in ork1 correlates with the failure of follicles to elongate along the a-p axis. Accordingly, it has been proposed that the planar-polarised bMF-pattern in wt provides a molecular corset restraining the increase in size along the transversal axis and contributing to follicle elongation [18,22]. It is known that Lar, a receptor tyrosine phosphatase, interacts with extracellular matrix proteins as well as with the bMF-cytoskeleton and is required for polarised bMF-organisation [18,20,21]. Consistent with this, mutants of LanA, a component of the extracellular matrix being polarised perpendicular to the a-p axis of the follicle, produce round eggs as well [21,22]. Moreover, a screen for round-egg mutants revealed a possible function of the Nuclear Dbf2-related (NDR) kinase Tricornered (Trc) in regulating either bMF, cell-extracellular matrix interactions or transcription-factor activity [50]. Trc and its activator Fry, and Msn, a presumed upstream activating kinase of Trc, are also required for planar cell polarity in the FE at early stages of follicle elongation [50].
In addition, cell-cell communication is needed for the planar polarisation of bMF in FC, since mutations in the atypical cadherin fat2 (allele of kugelei) show a particularly strong round-egg phenotype [18].
Considering that all these genes are part of a pathway establishing planar cell polarity in the FE, we assume a function for ork1 in the same pathway. Since planarpolarised bMF-orientation requires the orchestrated action of a large number of FC [18], the variability (within a follicle as well as between follicles) of bMF-orientation after soma-knockdown of ork1 seems reasonable. Consistent with the fact that the round-egg phenotype of ork1 is especially prominent in later developmental stages, the follicle-shape defects in mutants of trc, fry or kugelei are prominent not before S10 [22,50]. As consequence of the ork1-knockdown, we also observed a disturbed MT-alignment along the a-p axis as well as a weaker MT-cytoskeleton in general. In insects, a polarised MT-pattern in the FE has long been associated with the control of egg shape [70].
As current knowledge about planar FC polarity and follicle elongation comes predominantly from the described round-egg mutants, the involvement of ion-transport mechanisms, like Ork1, adds new insight into these processes. A role for Nhe2 in Fz-mediated planar cell polarity signalling has already been reported [13]. The authors suggested a model in which Dsh binds weakly to Fz, and the proximity of Fz to Nhe2 helps to maintain a slightly basic local pH i which facilitates the interaction of Dsh and Fz. Under acidic conditions, however, this interaction is weakened, leading to a repulsion of Dsh from Fz [13]. Moreover, the V-ATPase-subunit VhaPRR has been identified as a regulator of Wingless and planar cell-polarity signalling: VhaPRR could promote a favourable pH i -environment that supports Fz-signalling, alters Fz-conformation, promotes assembly or regulates Fz-trafficking [29].
The V mem -and/or pH i -dependent binding and surface recruitment of signalling-pathway components is one possible way how bioelectrical signals, generated by iontransport mechanisms, exert influence on signalling pathways. Therefore, we propose that ork1, besides its other reported functions [71,72], is involved, via bioelectrical signalling, in the establishment of planar cell polarity in the FE, thereby contributing to an elongated follicle shape. This interpretation is in accordance with previous studies suggesting influences of V mem -and pH i -changes on cytoskeletal organisation and planar cell polarity [5,16]. Moreover, correlations between alterations in bioelectrical patterns and changes in planar cell polarity were recently described in the mutant gurken [17].

Conclusion
The genetic tool box of Drosophila provides several means for a refined and extended analysis of bioelectrical phenomena. Both the V mem -sensor ArcLight, initially designed to track action potentials in neurons, and the pH isensor pHluorin-Moesin, initially designed to analyse phagocytosis, are useful tools to investigate tissue-specific bioelectrical properties during oogenesis. In comparison to fluorescent indicator dyes, genetically-encoded sensors provide several technical and practical advantages. For some types of experiments, however, the use of indicator dyes appears more suitable, since they exhibit higher sensitivity to small bioelectrical changes. Similar to earlier experiments using inhibitors, the modulation of bioelectrical signals via RNAi-knockdown of genes coding for ion-transport mechanisms and gap-junction subunits resulted in distinct cytoskeletal changes. Moreover, RNAi-knockdown exerted influence on crucial processes during development of the ovary and oogenesis. Therefore, by using genetic tools, further evidence amounts for bioelectrical regulation of developmental processes via control of both signalling pathways and cytoskeletal organisation.

Preparation of follicles
Female flies were killed, 2-3 days old ovaries were dissected, and single follicles of vitellogenic stages S8-S12 were isolated as described previously [5,16,17]. Dissection and cytoskeletal staining were carried out in Drosophila phosphate buffered saline [75], whereas inhibition experiments and morphological analysis were carried out in R-14 Medium [75,76].

Optical sectioning of living follicles
Single follicles of S8-S12, expressing either ArcLight or pHluorin-Moesin, were imaged in R-14 medium on a Zeiss AxioImager.M2 structured illumination microscope (SIM), equipped with a Zeiss ApoTome and a Zeiss AxioCamMRm camera, using a × 20/0.5 objective. Median optical sections were produced, and ImageJ (NIH, USA) was used to generate pseudocolour images as described previously [5,16,17].

Inhibition experiments
All S10B-follicles of a single fly (approximately 10-20), expressing either ArcLight or pHluorin-Moesin, were divided into a control group and an experimental group. Inhibition was performed for 20 min in R-14 medium containing one of the following inhibitors of iontransport mechanisms (cf. [5,16]): Na + /H + -exchangers (NHE) and amiloride-sensitive Na + -channels were blocked with amiloride (Sigma-Aldrich, Germany; 1 mM; dissolved in dimethyl sulfoxide; DMSO), V-ATPases were blocked with concanamycin A (Biomol, Germany; 1 or 2.5 μM; dissolved in DMSO), ATP-sensitive K +channels were blocked with glibenclamide (Biomol; 250 μM; dissolved in DMSO), voltage-dependent L-type Ca 2+ -channels were blocked with verapamil-HCl (Sigma-Aldrich; 1 mM; dissolved in 70% ethanol), Cl − -channels were blocked with 9-anthroic acid (Sigma-Aldrich; 1 mM; dissolved in DMSO), and Na + /K + /2Cl − -cotransporters were blocked with furosemide (Sigma-Aldrich; 1.5 mM; dissolved in DMSO), respectively. R-14 medium containing 0.25-1% v/v ethanol or DMSO was used in control experiments. Immediately after incubation, groups of three to seven follicles were imaged in covered glass block dishes on a Zeiss Axiovert 200 wide-field fluorescence microscope (WFM), equipped with a Hamamatsu Orca ER camera, using a × 10 objective as described previously [16]. During the respective experiments with either the V mem -or the pH i -sensor, exposure time and other settings remained unchanged.

Quantification of fluorescence intensities
Original grey-scale WFM-images (Fig. 1d) were used to measure, with ImageJ, the fluorescence intensity ("mean grey value") in the columnar FE of each follicle. The values of control follicles were averaged; then values of control and treated follicles were normalised to the mean of the control group. For each inhibitor, the experiment was repeated at least four times. To consider the variability between follicles, all normalised values of the same treatment were averaged (relative intensity, Fig.  2). To consider the variability between experiments, for each treatment a mean intensity ratio (Figs. 3 and 4) of the mean values of the experimental and the control groups (inhibited/control) of four repetitions was calculated. The mean values were compared using either an unpaired t-test (Fig. 2) or a one-sample t-test (Figs. 3  and 4). Microsoft Excel and GraphPad Prism were used for statistical analysis, and GraphPad Prism was used for data presentation.

RNAi-knockdown screen
Candidate genes of ion-transport mechanisms and gapjunction subunits showing enriched ovary expression (with respect to the signal in whole flies) were selected according to the FlyAtlas 2 Gene Expression Database (http://flyatlas.gla.ac.uk/; cf. [52]). In a first experiment, the respective VDRC UAS-strains were used for RNAi in the FE. Since these RNAi-constructs (with the exception of #v46553) had no effects on either ovary morphology or oogenesis (for summary, see Additional file: Table S1), the screen was repeated using TRiP UAS-lhRNAand UAS-shRNA-strains. Short hairpins (sh) embedded into a micro-RNA backbone are known to be very effective for knockdown in both germline and soma [74]. Males of the driver lines mat-tub-Gal4-GeneSwitch [58], mat-tub-Gal4.VP16, MTD-Gal4 or tj-Gal4 were crossed with UAS-lhRNA or UAS-shRNA females (in Valium10 vector [68] for soma knockdown, or in the very effective Valium20 vector [74] for soma and germline knockdown). F1 females, reared at 25°C for 3 days on standard medium with additional fresh yeast, were dissected (n ≥ 10 flies for each strain). Ovaries from transcriptionally almost inactive mat-tub-Gal4-GeneSwitch > UAS-lhRNA flies or mat-tub-Gal4-GeneSwitch > UAS-shRNA flies were used as controls.

Staining of microtubules
Follicles of wt and ork1-knockdown (BL53994) flies were fixed, incubated with a monoclonal antibody against acetylated α-tubulin (6-11B-1; Santa Cruz Biotechnology, USA), and stained as described in detail previously [5,17]. Thereafter, the follicles were mounted and analysed as described above using tangential optical sections (n = 18 ork1-follicles). Control follicles were treated without primary antibody.
Additional file 1: Table S1. Summary of candidate genes showing no effects in RNAi-knockdown screen. Data corresponding to Table 1.