Gradual segregation of Adult Stem Cells and Niche cells during development from common precursors under the guidance of graded extracellular signals

Adult stem cell function relies on the prior specification and organization of appropriate numbers of stem cells and supportive niche cells during development. Insights into those developmental processes could facilitate the synthesis of organoid mimics. Drosophila Follicle Stem Cells (FSCs) present an amenable paradigm with many similarities to mammalian gut stem cells. In an adult germarium a central domain of about 16 FSCs produces a posterior stream of transit-amplifying Follicle Cells (FCs), which encapsulate mature germline cysts to support egg development and, from their anterior face, quiescent Escort Cells (EC), which support the maturation of germline cysts. The behavior of FSCs is guided in part by niche signals produced by ECs and by a specialized polar cell FC derivative. Thus, ECs and FCs are both adult stem cell products and niche cells. Here we show by lineage analyses that adult ECs, FSCs and FCs derive from common precursors during pupal development. We infer that disparities in initial anterior-posterior (AP) precursor location followed by limited dispersal of progeny leads to a gradual acquisition of distinct fates determined by final AP location, with progeny of a single precursor commonly straddling EC and FSC, FSC and FC, or all three territories through most of pupal development. Direct visualization of pupal ovaries, including live imaging revealed a transient population of FC precursors posterior to the developing germarium awaiting emergence of the most mature germline cyst. The consequent budding process was quite different from the budding of egg chambers in adults. An anterior to posterior gradient of Wnt signaling develops shortly after pupariation. Loss of pathway activity cell autonomously resulted in more posterior adult progeny fates, while increased pathway activity elicited the opposite response, suggesting that stronger Wnt signaling favors anterior migration. Clearly detectable JAK-STAT pathway activity emerges only halfway through pupation after polar cells form, and spreads from posterior to anterior. Loss of JAK-STAT activity had similar consequences to increased Wnt pathway activity, drastically reducing FSC and FC production cell autonomously, while increased JAK-STAT activity promoted excessive precursor proliferation. We conclude that FSCs develop in co-ordination with their niche and product cells, that specification of stem cell identity is gradual, subject to stochastic influences and guided by graded extracellular signals, presaging similar regulation of adult stem cell behavior by the same pathways.


Introduction
Adult stem cells must be established in suitable numbers and niche environments during development to ensure that their activity is properly regulated and suffices to maintain the relevant adult tissue. Identifying stem and niche cell precursors and ascertaining how their coordinated development is organized would provide an essential foundation for understanding how the normal process may go awry, potentially conferring a subsequent deficiency in tissue repair or cancer susceptibility. In addition, it would facilitate the creation ex vivo of functional stem cell units as organoids from pluripotent stem cells, which can typically only be guided in isolation towards immature (non-adult) cell fates.
The development of most terminally differentiated adult cell types generally results from a series of successive definitive and irreversible decisions that are stabilized by potentially long-lasting shifts in gene expression and chromatin modification [citations]. The development of adult stem cells has not been studied extensively and may differ significantly from that paradigm because stem cells typically continue to proliferate in adulthood and often exhibit malleable relationships with their products, with the potential for marked changes in interconversion rates in either direction.
Drosophila ovarian Follicle Stem Cells (FSCs) present a particularly interesting paradigm because the organization, behavior and regulation of these adult stem cells has been studied quite thoroughly to reveal many similarities to mammalian gut stem cells as well as a variety of interesting, potentially prototypical features (Gehart and Clevers, 2019;Hayashi et al., 2020;Reilein et al., 2017).
Notably, FSCs are maintained by a form of population asymmetry wherein division and differentiation are not generally coupled (Reilein et al., 2018b), FSCs directly produce two instantaneously distinct types of cells, both of which produce important niche signals, and the proliferation rates and immediate differentiation potential of FSCs is heterogeneous but spatially organized. All of these behaviors are largely regulated by extracellular signals that are graded along the major (anterior-posterior) physiological axis (Reilein et al., 2017;Vied and Kalderon, 2009). FSCs are also of special interest because they must coordinate with germline cells, fueled by nearby Germline Stem Cells (GSCs) to produce functional products during oogenesis (Hayashi et al., 2020;Kahney et al., 2019;Waghmare and Page-McCaw, 2018). The mechanisms of coordination in the adult are currently not well understood but must rely on an organization that is established during development (Gilboa, 2015).
In the adult ovary, GSCs and FSCs are housed in the most anterior region of each of the 30-40 ovarioles, known as the germarium (Fig. 1J) (Hayashi et al., 2020). Somatic Terminal Filament (TF) and Cap Cells at the anterior of the germarium are stable sources of niche signals for GSCs and FSCs, with GSCs directly contacting Cap Cells. GSC Cystoblast daughters divide four times with incomplete 6

Outline of ovary development
We dissected and imaged developing ovaries during pupation, highlighting germline cells with Vasa antibody, Terminal Filament and Cap cells with antibody to Lamin-C, and somatic ICs (and Cap Cells) with antibody to Traffic Jam (TJ) in order to provide a systematic descriptive timeline (Fig. 1). The germaria within each ovary develop at slightly different rates (from medial to lateral) and the overall duration of pupation can vary significantly, leading to some differences in the reported timing of critical stages (Cohen et al., 2002;Lai et al., 2017;Li et al., 2019). In common with other studies, we assign "0h APF" as the time when larvae cease moving (pupariation) and used that event to collect animals at known intervals after pupariation. Animals mostly eclosed five days later (120h APF). The first egg chamber had budded from about half of all germaria at about 56h APF (Fig. 1H).
Between the TFs and basal cells are developing germaria containing germline cells (green in Fig. 1) and somatic cells (white in Fig.1A-I). Six of these somatic cells surround the most posterior TF cell, the Transition Cell (Panchal et al., 2017), and become non-dividing Cap Cells (Fig. 1J), expressing Lamin-C and TJ shortly after the larval/pupal transition in response to Notch signaling and an ecdysone pulse (Gancz and Gilboa, 2013;Gancz et al., 2011;Panchal et al., 2017;Song et al., 2007;Yatsenko and Shcherbata, 2018). Primordial germline cells (PGCs) that contact Cap Cells largely remain in position to become adult Germline Stem Cells (GSCs) (Asaoka and Lin, 2004;Zhu and Xie, 2003). The remaining PGCs will develop similarly to adult Cystoblasts to produce 2-, 4-, 8-and finally 16-cell germarial cysts during pupation. The somatic cells posterior to Cap Cells in the developing germarium are interspersed amongst germline cells and were therefore named Intermingled Cells (ICs) (Fig. 1J).
These cells express TJ (white in Fig. 1A-I) but not Lamin C or Fas3 (both red in Fig. 1B-I) at pupariation, allowing a clear distinction from Cap Cells and basal stalk cells, respectively (Gilboa, 2015;Lai et al., 2017;Panchal et al., 2017). It has generally been assumed that ICs are the source of adult ECs (Gilboa, 2015) but neither this relationship nor the source of FSCs and FCs in newly-eclosed adults has been defined experimentally. Here, we use lineage analyses and the changing morphology of developing ovarioles during pupation to investigate the origin of adult somatic stem and niche cells.

Spatially organized progression of germline differentiation
The progress of germline development was monitored by by using Hts antibody (red in Fig. 2) to reveal a rounded spectrosome in single cells or the branched shape of the fusome, which connects Vasa-marked germline cells in a cyst (green in Fig. 2), and by incubating samples with EdU to assess DNA replication as a proxy for cell division (Fig. S1). Primordial Germ Cells (PGCs) do not initiate differentiation before late larval stages due to the repressive action of ligand-free ecdysone receptor and Broad transcription factor targets in somatic cells (Gancz et al., 2011). Thereafter, anterior PGCs receive Dpp signals from newly-formed Cap cells that prevent differentiation, allowing them to become adult GSCs, while posterior PGCs start to express the early differentiation marker, Bam-GFP in wandering third instar larvae just prior to pupariation, co-incident with a strong ecdysone pulse (Gancz et al., 2011;Zhu and Xie, 2003). A short time later, some branched fusomes are seen in posterior PGCs, indicating incomplete divisions with incomplete cytokinesis to form cysts with 2 or more cells (Gancz et al., 2011;Li et al., 2019). The progression of germline differentiation initiated in posterior PGCs during pupation has not been described.
We observed 2-cell cysts at 0h APF, cysts of up to 8 cells by 12h APF and sixteen cell cysts at 18h APF ( Fig. 2A-C). Progressively larger cysts were arranged, with few exceptions, from anterior to posterior from 18h APF onwards; this was particularly clear in samples from 24h to 48h APF ( Fig. 2D-F). Individual germaria within a developing ovary exhibited different degrees of progression, together with AP elongation of the germarium to accommodate increasingly large germ cell units, and in some cases germline cells had migrated posteriorly out of the germarial region ( Fig. 2A-D). Multiple sixteen cell cysts were seen in individual germaria by 36h APF. The most posterior location included two cysts side by side in germaria at 24h APF but only a single cyst occupied that position in most cases by 36h APF and in all germaria by 48h APF (Fig. 2D-F). Consistent with these observations of cyst progression and location, EdU labeling was often seen in the most posterior cyst(s) up to 24h but rarely at 36h, while at 48h even the second most posterior cyst in the germarium rarely incorporated EdU 8 ( Fig. S1A-D). Our observations reveal a progression of germline differentiation that is spatially organized along the AP axis during the first 48h APF, at which time the organization of germline cysts resembles that in the adult up to a stage 2b cyst (Fig. 1J).

Follicle Stem Cells and Escort Cells derive from a shared precursor
We used lineage analysis to learn about the specification of ECs, FSCs and FCs from precursors during pupal development. We chose MARCM clonal analysis (Lee and Luo, 2001) because positive cell marking greatly facilitates scoring all cells in a clone. We used a mild heat-shock to induce clones at defined times via a heat-shock inducible Flp recombinase transgene (hs-flp) to target all dividing precursors without bias and to try to produce clones derived from a single marked cell per ovariole. Maturation of a strong GFP MARCM signal in adult and developing ovarioles requires almost 4d. We therefore dissected ovarioles from adults 2d after eclosion ("+2d"), having induced clones daily (in separate experiments) between 6d and 2d before eclosion ("-6d" to "-2d"). Each ovariole was scored for the number of FSCs (defined as within the three layers of somatic cells anterior to the border of Fas3 staining) (Fig. 1J), the number of ECs (classified as region 1 or 2a, according to AP location) and for the presence of marked FCs in the germarium and first two egg chambers, by imaging and archiving complete confocal z-stacks (Fig. 3). Marked TF and Cap Cells were also noted. Clones were classified according to inferred cell types present at the time of eclosion (see Methods for extrapolating data from +2d adults to 0d adults).
The average number of labeled ECs and FSCs in a marked adult germarium decreased when clones were induced at progressively later stages (from an average of 8.1 at -6d to 2.0 at -2d) and very few marked clones were observed in the absence of heat-shock, consistent with almost all clones arising at the time of deliberately administered heat-shock and continued division of precursors through pupal development. As expected, TF cells were almost never labeled and CC labeling was seen only at the earliest times of induction (3% of clones at -5d) because these cells have already differentiated shortly before (TFs) or just after (CCs) the larval/pupal transition and the MARCM method only labels dividing cells.
Strikingly, almost all early-induced clones (from -6d to -4d) that included marked FSCs also contained marked ECs ("EC/FSC clones") ( Fig. 3A, D-G) (Table 1). Clones containing only marked ECs ("EC only") were seen at each time of clone induction, while a high frequency of clones with marked FSCs but no marked ECs ("FSC only") was evident only when induced at -3d or -2d (Fig. 3A, K). These observations immediately suggested that most adult FSCs derive from a dividing precursor that gives 9 rise to both ECs and FSCs. During the second half of pupation an increasing proportion of the products of early EC/FSC precursors produce FSCs but no ECs.

Relative location of precursors along AP axis instructs fate
We found that marked cells in an adult germarium were generally clustered in the AP dimension, presumably centered on AP locations determined by the position of the first marked progenitor cell. This can be seen for all times of clone induction tested, from individual images ( Fig. 3D-K) and also numerically after classifying locations more finely by counting marked region 1 (r1; most anterior) and region 2a (r2a) ECs separately. Of all clones (induced from -6d to -2d) containing more than one cell and at least one r1 EC, 35% contained no other cell type, 19% included also only r2a ECs, while only 8% also included only FSCs. Thus, sisters of a region 1 EC are most commonly only in region 1; they are also more commonly found in region 2a than in the more posterior FSC region, even though the total frequency of clones with region 2a ECs (54%) and FSCs (57%) were similar (Fig. 3B).
Conversely, of all clones (induced from -6d to -2d) containing at least one FSC, 25% contained no ECs, 18% included also only r2a ECs, while only 11% also included only r1 ECs, even though the total frequency of clones with r1 ECs (68%) was higher than for r2a ECs (54%) (Fig. 3C). These observations support the hypothesis that precursors occupy significantly different AP positions throughout pupal development, their progeny undergo stochastic movement in either direction but still remain relatively close in the AP dimension and cell identity (r1 EC, r2a EC or FSC) is set by their AP location at eclosion. By adulthood this identity includes starkly different behaviors of quiescence (ECs) or proliferation (FSCs).

EC precursor division declines from the anterior at mid-pupation
For each time of clone induction, we summed the total number of marked FSCs and ECs (split into r1 and r2a) over all clones (Table 2). The ratio of marked ECs (r1 or r2a) to FSCs reports the relative number of the different cell types subsequently produced from dividing precursors. The absolute number of ECs to be made from dividing precursors in a single germarium was then calculated by assuming that each adult germarium has sixteen FSCs and that FSC precursors, like FSCs themselves, do not cease dividing (hence, prior to eclosion there are always 16 adult FSCs to be made from dividing precursors) (Table 2). Importantly, this calculation is based on scoring all labeled cells over all clones and is therefore equally valid whether the scored cells in an ovariole derive from a single marked cell or from more than one marked cell.
The number of ECs in an adult germarium was found to be about 40 on average but was quite variable among individuals of a single genotype (32-51 over 25 germaria), as noted previously (Wang and Page-McCaw, 2018). At -6d and -5d the inferred number of ECs still to be made from dividing precursors was almost identical at 33 (19 r1 and 14 r2a) per germarium (Table 2). This approximates the number of adult ECs, consistent with the idea that all or almost all cells that will become adult ECs are still dividing at the start of pupation. The inferred number of ECs still to be made from dividing precursors dropped slightly at -4d (from 33 to 28) and more markedly by -3d (to 15) and -2d (to 13) ( Table 2). We surmise that precursors of over half of the future adult EC population have ceased dividing by 2-3 days prior to eclosion. Moreover, the average number of dividing precursors for r1 ECs (4.7) was significantly lower than for r2a ECs (8.6) by -2d even though the total adult r1 EC population is higher (19 vs 14, inferred from -6d and -5d data). Thus, a reduction in precursor division appears to spread from the anterior towards the posterior as pupal development proceeds.

Low frequency clones to validate a common precursor for ECs and FSCs
The overall clone frequency (percentage of ovarioles with any marked cells in the germarium or early egg chambers) in the MARCM experiments tracing the fate of precursors marked from -6d to -2d was 60%, yielding an expectation that about 2/3 of marked ovarioles derive from marking a single precursor if all recombination events are independent and equally likely (see Methods). To obtain a lower frequency of marked clones, and hence a greater proportion deriving from a single cell, we used a multicolor marking strategy wherein a recombination event can lead to loss of GFP or a lacZ transgene, with or without RFP loss to yield six possible phenotypes (Reilein et al., 2017). Hence, for a given frequency of clone induction the proportion of a specific color combination will be substantially reduced. The two types of clone with the lowest frequency (because they require recombination at FRT40A and FRT42B) were GFP-only and lacZ-only, but we scored only the former because it was sometimes difficult to score all marked ECs for lacZ-only clones. GFP-only clones, induced at -5d and scored in newly-eclosed (0d) adults, were found at an overall frequency of 26/115 ovarioles so that an estimated 86%, calculated as for MARCM clones (see Methods), would be expected to derive from a single marked cell.
The patterns of GFP-only and other multicolor clones were very similar to those observed by MARCM analysis (Fig. 4). Among GFP-only clones that included FSCs, eight also included ECs and only two did not (Fig. 4B, C). We can therefore be confident that a single precursor at -5d can give rise to both ECs and FSCs and that most FSCs originate from such precursors. The proportions (and yields) of GFP-only clone types were 42% EC-only (average 3.0 ECs), 31% EC/FSC (average 3.5 EC and 2.6 FSCs) and 8% FSC-only (average 4.0 FSCs) (Table 3).
Multicolor lineage tracing also showed clustering along the AP axis for individual GFP-only clones ( Fig. 4A-C). Additionally, clones of different colors in the same germarium were generally offset along the AP axis. In some cases, the two clones likely derive from a single recombination event in a common parent, yielding "twin-spot" daughters ( Fig. 4D-F). These observations graphically confirm the earlier conclusion that marked lineages spread a limited but significant distance along the AP axis (even for the two daughters of a cell undergoing recombination) and that the initially marked precursors have significantly different AP locations.

Follicle Cells produced during pupation derive from precursors that also yield FSCs and ECs
In the MARCM analysis presented earlier we inferred cell distributions in newly-eclosed adults from observations made 2d later in order to allow at least 4d for GFP marking to become clear. For MARCM clones induced early (at -5d) we were also able to examine newly-eclosed adults directly. This allowed us to score all marked FCs produced from precursors, as well as FSCs and ECs, because no egg chambers have matured beyond the point where FCs die and even the most recently produced FCs in region 2b of the adult germarium are readily distinguished from FSCs by Fas3 expression. From these data we can learn about the developmental origin of FCs, which are typically present surrounding two cysts in the germarium and four budded egg chambers in a newly-eclosed adult.
Most clones with marked FSCs also had marked ECs (24/27) (Fig. 5A, D), as observed in MARCM analysis at +2d and multicolor clone analysis, confirming a common precursor of ECs and FSCs. Almost all clones with marked FSCs also had marked FCs (25/27) (Fig. 5A), indicating a common precursor for adult FSCs and FCs. A significant proportion of all clones (23%) had marked ECs, FSCs and FCs, indicating that some individual pupal precursors give rise to all three cell types in newly-eclosed adults. Many ovarioles ("FC-only"; 25%) also contained marked FCs but no marked FSCs (Fig. 5C, D).
The locations of marked FCs were scored as "immediate" (region 2b), "germarial" (region 3) or in egg chambers ("ECh") 1-4 (occasionally three or five egg chambers were present). The percentage of clones with marked immediate FCs was higher for FSC/FC clones (56%) than for clones with marked FCs but no marked FSC ("FC-only") (16%). Conversely, the percentage of clones with marked FCs in at least one of the two terminal (most posterior) egg chambers was higher for FC-only clones (88%) than for FSC/FC clones (67%). Indeed, occupancy by marked FCs decreased gradually from anterior to posterior for FSC/FC clones and with the opposite polarity for FC-only clones (Fig. 6A). Thus, derivatives of a given precursor were clustered along the AP axis in FSC/FC territory, as previously noted for EC/FSC territory. The idea that the fate of precursor derivatives is set largely by the AP position of precursors, with limited AP dispersal of derivatives, appears to be the governing principle all the way from the most anterior r1 EC derivatives to the most posterior FCs in newly-eclosed adults.
Notably, while some clones contained only ECs or only FCs, clones with marked FSCs always contained ECs or FCs, indicating that there were no FSC-specific precursors at this time and that individual lineages regularly spanned the borders of adult cell identities.

Improved estimates of numbers of different precursor types just prior to pupation
Several ovarioles from animals heat-shocked 5d prior to eclosion contained marked FCs and marked ECs but no marked FSCs, strongly suggesting that these clones arose from two cells (because they were, unlike other clones, discontinuous). The frequency of such ovarioles (16%) was significantly higher than expected if all recombination events were independent (80% of ovarioles were predicted to contain single lineages because 67% of ovarioles contained no marked cells; see Methods). It is indeed possible that non-uniform temperature changes during heat-shock or other factors lead to some degree of clustering of recombination events rather than an even distribution among all ovarioles and all flies, especially when using mild heat-shock conditions to induce recombination events at a very low frequency. To account for the observed proportion of ovarioles with ECs and FCs we estimated that 30% of ovarioles with marked cells derive from a single marked cell (rather than 80% as estimated from the frequency of unmarked ovarioles), while the other 70% derive from exactly two marked cells. Based on this assumption, we calculated the component single cell clone frequencies and component cell numbers that would give rise to the scored raw data (see Methods). With this revision, the estimated number of precursors just prior to pupariation (see below) for precursors giving rise only to ECs (15.1), to EC and FSCs (with or without FCs) (3.1), only FSCs (with or without FCs) (1.2) or only FCs (4.6) totaled about 24 ( Fig. 5E; Table 3). The number of TJ-positive cells counted in pupal ovaries at 12h APF was 38, roughly consistent with the revised estimates from lineage analysis.
Multicolor GFP-only clones were used previously to provide the most definitive evidence that an individual precursor cell can produce both adult ECs and FSCs; we cited an expected frequency of 86% of marked ovarioles deriving from a single cell if recombination events were random and independent (77% had no marked cell). The observation that 42% of marked ovarioles included only ECs while 8% included marked ECs and FCs suggests that a slightly lower proportion of ovarioles harbor single-cell lineages. If we assume that only 70% of ovarioles with marked cells derive from a single cell, with the 13 remaining 30% deriving from two cells (likely a small overestimate of double-clones) we can calculate the component single cell lineage frequencies and component cell numbers that would give rise to the scored raw data, as above (see Methods). With this revision, we still find evidence that a single precursor cell gives rise to both ECs and FSCs (25% of all marked lineages) and that the frequency of such precursors greatly exceeds precursors that produce FSCs but no ECs (10%). The deduced proportions of clone types and the number of precursors of each type (9.0 EC, 4.2 EC/FSC, 1.6 FSC, 2.4 FC and 17 total precursors) are similar to those calculated above from the similarly timed MARCM experiment ( Fig. 5E; Table 3).
We used the same methods to estimate single-cell lineage types and yields from a MARCM analysis of clones induced 3.5d prior to eclosion. Here the proportion of labeled ovarioles (94%) and of EC/FC clones was particularly high, probably because there are many more dividing precursors than at -5d, so the deduction of single cell lineage content is likely less reliable. In both the raw data and derived lineage types it was clear that the proportion of FC-restricted lineages was much higher than at earlier times, while marked FSCs were still mostly accompanied by marked ECs, consistent with our earlier analyses of MARCM clones analyzed in in 2d-old adults. We estimated that the proportions of single-cell lineage types were EC-only (51%), EC/FSC(/FC) (7%), FSC(/FC) (4%), FC-only (38%) and that the total number of dividing precursors was about 35 ( Fig. 5E; Table 3). Based on earlier analysis of MARCM lineages induced at -4d and -3d and examined at +2d, there should be between 5 and 18 non-dividing precursors (say 11.5) at -3.5d (Table 2). Thus, we estimate that there are 13.3 FC-only precursors within a total precursor population of 46.4 (29%) at this time (corresponding to 36h APF; see later).

FC founders per egg chamber and timing of FC recruitment to first egg chamber
During adult oogenesis, an FC is defined by continued association with a germline cyst. The first FCs are permanently recruited to germarial stage 2b cysts and it has been estimated that there are on average 5-6 founder FCs per cyst (Reilein et al., 2017;Reilein et al., 2018b). The number of founders can be determined by measuring the fractional contribution of a single founder to the whole FC epithelium around a single egg chamber. However, marked lineages induced in precursors to founder FCs will sometimes first divide and then contribute more than one founder FC to a given egg chamber.
We examined the distribution of marked FC contributions from FC-only precursors during pupation in order to gain insight into the number of founder FCs seeding each egg chamber and the time at which founder FCs were first restricted to populating only the first budded egg chamber.
When plotted in bins of 3% for clones induced at -5d, the distribution of marked cells as a percentage of an entire egg chamber FC epithelium are clustered around 9-12% (a peak with a central cavity) and 24-27% (Fig. 6B). The distribution observed from a similar analysis of clones induced at 3.5d before eclosion was extremely similar (Fig. 6D). We do not expect to observe sharp peaks in the distribution of FC contributions because each founder FC may subsequently divide a different number of times; in particular, polar and stalk cells cease division much earlier than remaining FCs (Margolis and Spradling, 1995;Tworoger et al., 1999), likely accounting for the high representation of contributions from 0-6%. The values clustered around 9-12% and 24-27% plausibly correspond to the most common contributions of a single founder cell and two founder cells, respectively. If we consider only FC contributions lower than 25.5% (the mid-point of the second peak) we will most likely include the majority of clones deriving from single FC founders and only a small number of clones deriving from two or more marked FCs; this subset can give an approximation (likely a modest overestimate) of the contribution from a single founder FC. Within this restricted sample, the average contribution over all egg chambers in newly-eclosed adults was 12.2% (33 FC patches) for -5d clones and 12.4% (43 FC patches) for -3.5d clones, suggesting an average of at least eight FC founders per egg chamber.
Considering only the terminal egg chamber, average FC occupancy was 12.2% for -5d clones and 11.9% for -3.5d clones, almost identical to the overall average.
To determine when FCs were first allocated to the most mature germline cyst during pupation we looked at the distribution of FC-only clones amongst egg chambers. For clones induced 5d before eclosion there were fifteen ovarioles with marked FCs in the most posterior egg chamber. Twelve of these also had marked FCs in more anterior regions, mostly occupying at least two additional egg chambers. Hence, FC precursors were not yet constrained to contribute to a single egg chamber. The distribution of FCs from FC-only clones induced 3.5d before eclosion was more restricted; 21 of 34 ovarioles with labeled FCs in the terminal egg chamber had no other labeled FCs (Fig. 6E, F). The 13 exceptions all had marked FCs in egg chamber 3 (Fig. 6E, G) and are too many to be explained by the presence of two lineages. In most ovarioles with marked FCs in egg chambers 3 and 4 (9/13) the FC contribution to egg chamber 4 is less than 25.5%, suggesting that the initially labeled precursor at 36h APF generally divided one time before one daughter became an FC founder for egg chamber 4 and the other later contributed to egg chamber 3. In the majority of ovarioles with only egg chamber 4 labeled (12/21), marked FCs occupied more than 25.5% of the epithelium, suggesting that in most cases the initially labeled precursor divided before contributing two founder FCs to egg chamber 4. Thus, it appears that founder FCs were fully allocated to the first-produced egg chamber roughly one cell cycle after 36h APF, by which time many of the first founders allocated had already divided. We do not know the cell cycle length for these FC-only precursors but expect (from lineage amplification data) that it is shorter than 24h and may be close to 11h, the estimated cycling time for early FCs in adult ovarioles (Margolis and Spradling, 1995).
Marked FCs occupied only the two terminal egg chambers (one or both) in 36 out of 52 ovarioles with FC-only clones induced -3.5d before eclosion (Fig. 6E). Marked FCs occupied only territory anterior to the two terminal egg chambers in six ovarioles. Some or all of the remaining 10 ovarioles with hybrid patterns plausibly harbor two FC-only lineages. Thus, by 36h APF the FC-only precursors appear to be segregated into two populations with the majority (roughly 46/62, or over 70%, if we consider the 10 hybrid patterns to contain one lineage of each type) contributing only to the two terminal egg chambers and other precursors contributing only to more anterior egg chambers.
Ovarioles that include marked FSCs only rarely contributed to the two terminal egg chambers (Fig. 6C).
Indeed, the few exceptions may well derive from ovarioles that harbor an FC-only lineage (53% frequency overall) in addition to an FSC-containing lineage (26% overall). Thus, at 36h APF (see below), it appears that over two-thirds of FC-only precursors are constrained to contribute to one or both of the first two egg chambers, while more anterior egg chambers are later populated by a combination of FC-only and FSC/FC precursors. It is likely that all FCs contributed to the two terminal egg chambers derive from FC-only precursors.

Integration of cell lineage and morphological evidence
In our cell lineage studies a "-5d sample" was generated by collecting very young adults (less than a 6h range) centered around 120h after heat-shock. The time between puparium formation and eclosion was generally a few hours less for these adults than for animals used in morphological studies but there is also a short delay between heat-shock and genetic marking of a cell. (Golic and Lindquist, 1989). Consequently, the time a marked cell is born in a "-5d sample" likely corresponds closely to puparium formation, consistent with the observed labeling of some Cap Cells but very few TF cells, which have ceased division by that stage. In our morphological studies, samples were initially collected relative to directly observed larval immobility of each animal analyzed. After this initial calibration we were able to determine pupal stage according to ovary morphology.
Lineage data suggested that some precursor lineages are restricted to the most advanced germline cyst at 36h APF (-3.5d), with the remainder following one division cycle later. By 48h APF, all germaria have a single cyst in the most posterior region (Fig. 2) but neighboring somatic cells only express very low levels of Fas3 (Fig. 7D). In adult germaria, even the most recently recruited FCs surrounding a stage 2b cyst express high levels of Fas3; a subset of those cells must then become specialized polar and stalk cells before the egg chamber can bud about 24h later. The ICs described above that are just starting to express Fas3 at 48h APF therefore appear to be far behind schedule, given that the first egg chamber will bud at about 56h APF on average (Fig. 1). These observations pose the question of whether these ICs are indeed the cells that will form the FCs of the first budded egg chamber.

An Extra-Germarial Crown (EGC) forms prior to egg chamber budding
Posterior to the intermingled germline cells and somatic ICs of the germarium are somatic cells with strong Fas3 staining at 24h APF onward ( Fig. 1; Fig. 7B). These have all previously been considered as progenitors of basal stalk cells and indeed most can be seen to be intercalating into a stalk structure from 36h APF onwards ( Fig.1; Fig. 7), as previously described (Godt and Laski, 1995;Vlachos et al., 2015). However, we noticed that the most anterior of these Fas3-positive cells express TJ and form a small multi-layered cone or crown around the posterior of the developing germarium that is quite prominent by 48h APF ( population of cells that expresses both TJ and Fas3 strongly is notable because there is evidence from manipulation of TJ activity in adult ovaries that TJ generally represses Fas3 expression (Li et al., 2003), while a principal characteristic that distinguishes adult FCs from other TJ-expressing cells (FSCs and ECs) is strong Fas3 expression (Reilein et al., 2017). Key questions, which we address below, are therefore whether the most posterior ICs with low Fas3 expression at 48h APF become the FCs of the first-produced egg chamber and what becomes of the newly-identified group of EGC cells.
A stage 3 germline cyst is initially separated from the posterior egg chamber by only a small number of somatic cells just prior to budding from an adult germarium. During budding, the cyst remains associated with adjacent somatic FCs, which emerge as a monolayer epithelium connected to the germarium by a thin, short stalk of specialized cells. By contrast, prior to budding of the first pupal egg chamber there is a large accumulation of somatic EGC cells posterior to the germarium and after budding, the first egg chamber is often found much further from the germarium than in adults (Fig. 1H).
The intervening space is occupied by partially intercalated TJ/Fas3-positive cells and by a small "secondary" EGC, crowning the germarium (Fig. 7G). In samples with two budded egg chambers (72-84h APF) or more, the newly budded egg chamber is close to the germarium (Fig. 1I), resembling adult ovarioles. Thus, formation of the first egg chamber involves a morphological process quite different from in adults; this is partly recapitulated by the second egg chamber before adopting a typical adult morphology.

Live imaging reveals EGC cells as a source of FCs for the first egg chamber
We adapted a live imaging protocol previously developed for adult ovarioles (Reilein et al., 2018a) in order to view pupal ovaries and track individual somatic cells. We were able to follow isolated live ovaries for up to 15h, during which time many cell movements and cell divisions were observed, suggestive of continued normal behavior. We first used samples where all cells were labeled by expression of a ubi-GFP transgene. Later, we used animals with multicolor clones induced 2-3d earlier, so that some cells had lost either GFP or RFP expression. Variability in the levels of GFP and RFP simplified the task of following individual cells over time.
From 30h APF a variety of tracked ICs mostly maintained their relative AP locations within the developing germarium but with a few exceptions (the cyan cell moved anterior to the pink cell in Fig.   8A). From 40h APF one of two similarly placed germline cysts adopts the most posterior position in the germarium while somatic ICs appear to move independently of the cysts (the shaded white cyst moved past the pink and white cells in Fig. 8B).
Imaging of an ovary from 48h APF onwards captured the key process of egg chamber budding for the first time ( Fig. 9; Fig. S2, S3; Movie S1). The most posterior germline cyst moved away from the germarium into EGC territory, as EGC cells enveloped the cyst. The cyan cell started within the EGC, contacted the cyst prior to 8.3h and remained associated for the next 6.5h ( Fig. 9; Fig. S3). The white cell, which initially lay posterior to the morphologically defined EGC and initially (from comparison to fixed images) likely does not have strong TJ expression, entered EGC territory within 4h and became associated with the germline cyst by 12h ( Fig. 9; Fig. S3). The magenta cell, initially just posterior to the cyst, moved to the midpoint of the budded egg chamber ( Fig. 9; Fig. S2C). The red cell, which started a little further anterior, and would be considered an IC, maintained contact with the posterior-moving cyst and was last seen at the anterior aspect of that cyst, leaving ambiguous whether it remained associated or moved past the cyst to await the second cyst to bud ( Fig. 9; Fig. S2C). The yellow cell in the germarium above started from a similar location but ended up anterior to the budded cyst, forming part of a secondary EGC just beyond the germarium ( Fig. 9; Fig. S2B). Thus, over the nearly 15h of live imaging the newly-budded cyst acquired a set of closely-associated cells of epithelial appearance that derived from the EGC region, slightly posterior to the EGC region and the extreme posterior of the germarium at 48h APF. Imaging of later stages (below) suggests that some of the cells surrounding the budding cyst may be moving stochastically with an anterior trend, so it seems possible that cells initially derived from the germarium may end up anterior to the first budded egg chamber but most cells derived from the EGC, which are further posterior, likely remain on the budded cyst.
Imaging from about 60h APF, cells on the first budded egg chamber were seen to move anterior and lose contact (Fig. S4A). The most posterior cyst in the germarium of the same ovary rounded up and started to move into the secondary EGC, which included dividing cells (Fig. S4B, C). ICs anterior to that cyst moved further posterior as the germarium elongated and mostly appear poised to associate with later cysts (Fig. S4C).
In summary from all live imaging spanning 30h-68h APF, it seems likely that all TJ-positive cells observed in fixed samples (Fig. 7) retain TJ expression and become one of the adult TJ-expressing cell types, ECs, FSCs or FCs. Furthermore, the FCs surrounding the most mature egg chamber in newlyeclosed adults appear to derive solely or principally from the Extra-Germarial Crown of somatic cells that accumulates between 12h and 56h APF. The ICs that express Fas3 weakly at 48h APF probably contribute principally to the second, third or even later egg chambers formed during pupation, although it remains possible that some of the most posterior ICs may also associate with one of the first two cysts to bud.
Live imaging also afforded information about cell movements within the germarium, revealing apparently stochastic movements in all directions that sometimes altered the spacing of pairs of cells along the AP axis. These observations are consistent with the deductions from lineage analysis that somatic cells tend to remain in the same relative AP positions over a period but that a group of sister cells can spread over the AP axis.

Pattern of precursor divisions along the AP axis
The analysis of MARCM clones in 2d-old adults provided average numbers for the yield of ECs and FSCs per precursor of each type starting at different times from 6d to 2d before eclosion (Table 1).
The proportion of dividing precursors of each type is also indicated by the frequency of different clone types. From these data it is clear that there is a severe reduction in EC-only precursor division rates from -4d to -2d (24h to 72h APF) with the frequency of such clones falling from 58% and 57% at -5d and -4d to 30% at -3d and 14% at -2d. Moreover, among EC-only precursors the decline in division spreads from the anterior; the yield of r1 ECs relative to r2a ECs declines from 2.2 (-5d) and 2.6 (-4d) to 0.9 (-3d) and 0.8 (-2d). The yield per EC/FSC and FSC precursor is consistently higher than for EConly precursors at all times of clone induction, indicating higher division rates in more posterior locations. These yields do not include FCs and therefore account for only a fraction of EC/FSC and FSC precursor output, precluding numerical assessment of whether the increasing proliferation gradient from r1 ECs through EC/FSC precursors extends further posterior.
We incubated dissected pupal ovaries at various developmental stages with EdU in vitro over a 1h time period to detect cells in S phase. From 0-24h APF ICs throughout the germarium often incorporated EdU (Fg. S1A-C). By 48h APF, the prevalence of anterior EdU-labeled cells was quite low, with EGCs, by contrast, exhibiting frequent EdU signals (Fig. S1D). Even at 60h APF, however, an occasional IC in the anterior half of the germarium could be found with EdU label; none were evident at 96h APF (Fig. S1E, F). At 96h APF some cells anterior to strong Fas3 expression, which may be r2a EC or EC/FSC precursors, did still incorporate EdU.
We also used the FLY-FUCCI system to measure cell cycling. In this system, RFP-and GFPtagged proteins accumulate during the cell cycle under the influence of a chosen promoter, while degrons from CycB and E2F1 promote rapid degradation of RFP and GFP fusion proteins at the start of G1 and S phase, respectively (Zielke et al., 2014). While RFP and GFP are expected to be reliably absent during G1 and S phase, respectively, the speed of recovery of each signal during the next phase of the cell cycle depends on the strength of transcription of FLY-FUCCI components. We used UAS-driven FLY-FUCCI transgenes and found C587-GAL4 to be more effective than tj-GAL4 or act-GAL4 in producing GFP and RFP FUCCI signals in the developing somatic IC population.
In adult germaria all r1 ECs express only GFP ( domain has retracted anteriorly, roughly to territory anterior to strong Fas3 staining and the pattern of FUCCI labels is quite similar to adults, with anterior cells uniformly in a G1 or G0 state, followed by cells largely in G1 and G2 and then some cells in S phase also (Fig. 10F). Thus, direct observation of EdU incorporation and cell cycle markers confirms the progressive decline in anterior IC proliferation over time that was inferred from lineage analyses and suggests that the anterior to posterior gradient of proliferation stretches all the way to posterior ICs and EGC cells, which appear to cycle rapidly.

A putative FSC precursor marker yields lineages that also produce ECs and FCs
Recent single cell RNA sequencing of developing ovaries in late third instar larvae led to the discovery of a variety of genes with regional expression patterns (Slaidina et al., 2020). A set of GAL4 drivers, which partially reproduced the expression patterns of selected parent region-specific genes, were used in lineage tracing experiments to determine precursor-product relationships. A bond-GAL4 driver was used together with a temperature-sensitive GAL80 transgene to initiate GFP-marked lineages using standard G-trace reagents (Evans et al., 2009), keeping animals at 29C during larval stages, switching to 18C about one day after pupariation and scoring marked cells in 2d-old adults (Slaidina et al., 2020). Most of the labeled ovarioles (31/32) included marked FSCs (inferred from the presence of early marked FCs) or FCs (the fraction with FSCs was not explicitly documented) and over half of these (19/31; one has only marked ECs) also contained marked ECs. We used the same reagents to compare outcomes directly with our timed lineage studies that target all dividing precursors, scoring results in newly-eclosed adults so we could clearly resolve FSCs and FCs that are directly produced from precursors.
We found a high frequency of marked cells when animals were kept at 18C throughout (37 labeled ovarioles out of 108) with marked ECs and FSCs present in almost all cases (35/37) ( Fig. 11A-C). We saw no increase in the proportion of ovarioles with marked ECs or FSCs when larvae were raised at 29C and moved to 18C about 2d after pupariation (eclosing 6d later; 13 of 44 ovarioles labeled) or when animals were raised at 18C before shifting to 29C for the final 3d or 4d of pupation (23 of 88 ovarioles labeled) (Fig. 11D). In all cases the observed clones were dependent on bond-GAL4 (a much lower frequency of marked lineages, with different patterns of marked cells, was observed in parallel experiments with a Con-GAL4 driver). Because we observed no marked temperature dependence for these EC/FSC-containing clones we cannot define the time at which cells were first marked. However, we expect that most lineages were initiated close to the larval/pupal transition based on the lineages themselves, as described below.
The marked lineages were similar (Fig. 11A-D) to those we described earlier using hs-flp to initiate clones at pupariation (5d before eclosion), suggesting that they had been induced at a similar time. The majority of ovarioles from the bond-GAL4 experiments (considering all three temperature conditions together) included either only marked ECs (about 40%) or marked ECs and FSCs (about 50%, almost all together with FCs) (Fig. 11D). Only one out of 73 labeled ovarioles had marked FSCs but no marked ECs; that ovariole also included marked FCs. The number of labeled ECs and FSCs (when present) per germarium was consistently higher than we observed with hs-flp induced clones induced at pupariation (Fig. 11A-C; Table 3). Thus, if the two sets of clones were initiated at similar times most labeled ovarioles in the bond-GAL4 experiment must contain lineages derived from more than one precursor. We therefore derived the component single clone types (frequency and number of products) that would produce the observed bond-GAL4 > G-trace results if we assume that every ovariole has exactly two marked lineages (roughly accounting for the increased cell numbers seen; see Methods). The results are remarkably similar to the single clone types and content deduced for hs-flp induced clones 5d prior to eclosion, (Fig. 11E; Table 3). The most significant differences were a greater yield of ECs per EC-only lineage and fewer FC-only lineages, consistent with slightly earlier clone initiation in the bond-GAL4 experiments. It is therefore plausible (though not the only possible explanation of the primary data) that bond-GAL4 labeled precursors sporadically throughout the IC region shortly before pupariation. Consequently, in keeping with results from MARCM studies, the observed FSC-containing lineages and those inferred to derive from a single cell also included ECs in almost all cases (as well as FCs). Our results do not support the labeling by bond-GAL4 of a precursor that is specific to FSCs and not ECs.
bond gene expression was described as being patchy but distributed among all ICs at early third instar before concentrating in a region just posterior to the IC domain (Slaidina et al., 2020). The results described above are consistent with bond-GAL4 driven clone initiation in the former pattern. Our live imaging studies suggest that cells in the latter, posterior location may contribute to the EGC by 36-48h APF and then to FCs of the first budded egg chamber. When animals with bond-GAL4 and G-trace were maintained at 29C during the last 3d or 4d of pupation we observed an additional type of lineage, either in isolation or together with the types of marked cells described above (55/88 ovarioles; 0/108 when at 18C throughout) (Fig. 11G). Here, a few isolated FCs or small clusters in the most mature egg chamber, and no others, were labeled, suggesting several independent recently initiated lineages.
Some of the GFP-labeled FCs also expressed UAS-RFP, indicating current or recent bond-GAL4 activity (Fig. 11G). Surprisingly, the same type of clone was observed if animals moved from 29C to 18C about 2d after pupariation (30/44 ovarioles, with only 6 ovarioles also including RFP-expressing FCs) (Fig. 11F), suggesting that some of the FC-only clones confined to the first budded egg chamber may originate from posterior EGC cells prior to budding of the first egg chamber.

Evolution of the pattern of Wnt pathway activity in the FSC niche
Wnt pathway activity in adult germaria, measured with a Fz3-RFP reporter, declines from a high-level in r2a ECs to zero in the earliest FCs (Fig. 12I, L) (Reilein et al., 2017;Wang and Page-McCaw, 2014). Genetic manipulations that increased Wnt pathway activity in an adult FSC autonomously resulted in a more anterior location with eventual conversion of all such FSCs to ECs, while genetic loss of Wnt pathway activity autonomously reduced EC production and caused FSCs to accumulate in the most posterior layer of the normal FSC domain (Reilein et al., 2017). The sources of Wnt in the adult germarium include high levels of Wg and Wnt6 production in Cap cells and lower levels of Wnt 2 and Wnt4 in ECs (Forbes et al., 1996;Luo et al., 2007;Sahai-Hernandez and Nystul, 2013;Upadhyay et al., 2018;Waghmare and Page-McCaw, 2018;Wang and Page-McCaw, 2018).
In pupal germaria Fz3-RFP was first detected 27h APF and showed a marked increase in intensity between 36h and 48h ( Fig. 12B-G, J). At all stages Fz3-RFP was higher in anterior than posterior regions, consistent with the possibility that Cap Cells, which are specified at the anterior of the germarium very close to the start of pupation, and even developing EC precursors, may be major sources of Wnt ligand, as in the adult. At 48h APF the decline in Fz3-RFP was spread over the posterior two-thirds of the germarium (Fig. 12 G, J). By 60h APF the domain of consistently high Fz3-RFP expression occupied a larger anterior territory, extending roughly halfway to the region of strong Fas3 expression, followed by a steeper decline to background levels ( Fig. 12H, K). At all stages, the border of strong Fas3 expression coincided with very low levels of Wnt pathway reporter activity, suggesting the possibility that Wnt signaling may inhibit Fas3 expression.

Influence of Wnt pathway activity on precursor behavior
To determine the influence of Wnt pathway activity on the behavior of individual precursors during pupal development we repeated MARCM lineage studies with the recombination event that marks cells simultaneously creating a homozygous mutation that either inactivates (null arr mutation) or hyperactivates (strong axn or apc1 apc2 double mutation) the Wnt pathway. We performed two sets of experiments with arr mutations and one each for axn and apc1 apc2, inducing clones at 2-6d before eclosion and examining ovarioles in 2d-old adults. Outcomes were similar for each pair of tests, so we aggregated the results for the two experiments with reduced Wnt pathway activity (arr), the two with increased activity (axn, apc), and three controls ( Fig. 13A-F).
Loss of arr activity consistently led to a decreased frequency of EC production relative to FSCs, while increased Wnt pathway activity produced the opposite outcome. Thus, the percentage of all cells in marked clones that were FSCs was, on average from -6d to -2d, 42% for controls, 61% for arr and only 20% for axn/apc, with the same trend evident for each time-point (Fig. 13C). The proportion of all clonal derivatives that become r1 ECs was consistently lower than controls for arr mutant precursors at all times of clone induction from -6d to -2d (average 13% vs 32%) (Fig. 13A, D, E). By contrast, the proportion of r2a ECs was markedly lower for arr mutant precursors than controls only when clones were induced at -2d (Fig. 13B). These results suggest that loss of Wnt pathway activity results in posterior movement of precursors relative to wild-type neighbors at all times from -6d onward. At early stages this results in r1-only precursors transforming to r2a and FSC-producing locations. At intermediate times the increase of r2a precursors from r1 precursors is offset by conversion of r2a precursors to FSC fates, while in the last stages there is a net deficit in r2a EC production because the majority of actively dividing precursors are in regions that can only produce r2a ECs or FSCs (and FSC fate is favored by loss of Wnt pathway activity). Thus, normal Wnt pathway activity is important for limiting posterior movement of precursors over a wide domain throughout pupal ovary development.
Increased Wnt pathway activity consistently increased the proportion of r2a ECs in clones induced from -6d to -2d, while the proportion of FSCs was reduced (Fig. 13B, C, D, F). These results suggest that cells with increased Wnt pathway activity adopt more anterior positions within the domains that normally give rise to either ECs or FSCs. The proportion of r1 ECs produced by axn/apc mutants was almost identical to controls for the entire -6d to -2d induction period (Fig. 13A). Thus, the reduced levels of Wnt pathway activity observed in posterior regions of normal germaria that include EC/FSC and FSC precursors, is important for maintaining a sufficiently posterior location to become an adult FSC.
It is not straightforward to measure the effect of altered Wnt pathway activity on precursor division rates from lineage data for a number of reasons. First, the cellular yield per clone is dependent on clone induction frequencies, which can be quite variable even within a single experiment. Second, alteration of cell locations will affect yields because more anterior cells cease division earlier. Third, the yield of FCs cannot be scored (and will likely differ for mutations affecting AP location). Notwithstanding these reservations, the average number of cells in clones with altered Wnt pathway activity was quite similar to controls (4.8 cells per clone for controls, 3.6 for arr and 5.4 for axn/apc on average for -6d to -2d clones), suggesting that there was no drastic change in the rate of precursor division over most of pupal development. In adults, excessive Wnt pathway activity was found to reduce FSC division rate significantly, while loss of Wnt pathway activity had little effect.
We also induced MARCM clones with increased (axn mutation) or decreased Wnt activity (expression of dominant-negative TCF (van de Wetering et al., 2002)) at -5d and examined ovarioles in newly-eclosed adults, allowing us to measure the production of FCs in addition to ECs and FSCs.
Similar to the observations made for clones examined in 2d-old adults, increased Wnt signaling resulted in a loss of marked FSCs, an increase in marked r2a cells and no change in the proportion of marked cells in region 1 (Fig. 13G). Also, reduced Wnt signaling led to reduced production of r1 ECs and an increase in the proportion of r2a ECs (Fig. 13G, I, J). However, there was no increase in the proportion of marked FSCs, contrasting with results seen for arr mutations; that is likely because dominant-negative TCF produced a lesser reduction in Wnt pathway activity.
The effect of genetically altered Wnt pathway activity on FC production was assessed from the distribution of clone types in newly-eclosed adults. Increased Wnt pathway activity reduced the frequency of clones with FCs but no FSC (from 26% to 2%) and the total frequency of clones that included marked FCs (from 51% to 5%) (Fig.13H). Reduced Wnt pathway activity did not have a large effect on the frequency of clones with FCs but no FSC (19% vs control 26%) or the total frequency of clones that included marked FCs (41% vs control 51%), reducing both proportions.
The effect of altering Wnt pathway activity on the distribution of EC and FSC clone types was in accord with results described above from counting total production of r1 and r2a ECs and FSCs summed over all clones. Thus, increased Wnt activity resulted in a greater frequency of clones with ECs but no FSCs (91% vs 63% in controls) and a lower frequency of clones with FSCs (7% vs 28% in controls) but with little change in the frequency of clones with marked r1 ECs (82% vs 78% in controls).
Wnt pathway reduction slightly reduced the proportion of clones containing r1 ECs (from 78% to 67%); it did not reduce the frequency of clones with ECs but no FSCs (63% vs 63% in controls) or increase the proportion of clones with marked FSCs (26% vs 28% in controls).
Thus, increased Wnt pathway activity increased EC production (but not r1 EC production) and drastically decreased both FSC and FC production. Both reducing and eliminating Wnt pathway activity reduced r1 EC production but only loss of pathway activity reduced FSC production.

Influence of JAK-STAT pathway activity on precursor behavior
JAK-STAT pathway activity in adult germaria, measured by a reporter containing multiple STAT binding sites (stat-GFP) is graded from posterior to anterior over the FSC domain, with ligand emanating principally from polar follicle cells, which first differentiate shortly before egg chamber budding (Bach et al., 2007;Vied et al., 2012). Loss of JAK-STAT activity has been reported as greatly reducing FSC proliferation, while excess JAK-STAT activity has the opposite effect (Vied et al., 2012).
Additional roles of JAK-STAT signaling in adult FSCs are currently under investigation.
In pupal germaria, stat-GFP expression was detected only at very low or background levels at early stages (Fig. S5A). A clear signal was first evident in a graded pattern declining from the posterior of the germarium at 60h and that pattern was maintained with increasing intensity thereafter (Fig. S5B-D). The onset of this pattern is shortly after the first egg chamber buds and therefore consistent with ligand production from polar cells, as in the adult. Within egg chambers stat-GFP expression was strongest adjacent to polar cells. To determine when polar cells form, we stained ovaries with Castor. In adults Castor is initially expressed in all early FCs but is then lost in all FCs other than polar and stalk 25 cells (Dai et al., 2017). We found Castor was confined to polar and stalk cells after the first egg chamber budded (Fig. S5E).
We manipulated JAK-STAT pathway activity in MARCM clones induced by heat-shock 5d before eclosion and examined marked lineages in newly-eclosed adults. Loss of stat activity resulted in a distribution of clonal derivatives very similar to that described for axn mutations (Fig. 13K). Thus, the fraction of all labeled cells that were FSCs was greatly reduced (6% vs 30% control) but the fraction of cells that were r1 ECs was unchanged (49% vs 51% control) (Fig. 13G). Accordingly, the proportion of clones that included FSCs was reduced (6% vs 28% control), the proportion of clones with ECs but no FSCs was increased (from 63% to 91%) and the proportion of clones with r1 ECs was not increased (70% vs 78% control). Loss of stat activity also reduced the frequency of clones with FCs but no FSC (from 26% to 6%) and the total frequency of clones that included marked FCs (from 51% to 8%) (Fig.   13H). Thus, in the absence of JAK-STAT activity we deduce that more posterior precursors, which would otherwise give rise to FSCs and FCs moved significantly anterior at some stage(s) during pupal development. The average size of clones was not significantly altered for the two major clone categories: clones with ECs but no FSCs (336/129=2.6 vs 156/62=2.5 control) and clones with ECs and Increased JAK-STAT activity due to expression of excess JAK (UAS-Hop) resulted in a large increase in clone size, with many marked cells throughout the germarium (Fig. 13L, M). Also, there was ectopic Fas3 expression in marked cells anterior to the normal Fas3 border (Fig. 13L) and some germaria had distorted morphologies, precluding reliable scoring of cell types. Clearly the proliferation of pupal IC cells can be stimulated significantly by artificially increasing JAK-STAT activity.

Discussion
An important general objective, exemplified here by FSCs, is to describe the origin of adult stem cells and surrounding niche cells and understand how the number and organization of these cells is instructed during development. Our investigations of changing morphology, key marker expression and cell lineage reveal a gradual and flexible adoption of cell types according to AP location throughout pupal development, rather than a series of discrete definitive cell fate decisions (Fig. 14). Thus, adult EC niche cells and FSCs develop from a common group of dividing precursors (ICs) interspersed with developing germline cysts and starting immediately posterior to Cap Cells, which serve as adult niche cells for both GSCs and FSCs, and are already specified by the start of pupation. The most anterior ICs become r1 and r2a ECs and markedly attenuate division, starting from the most anterior locations, during the second half of pupation (Fig. 14). More posterior ICs initially produce both ECs (mainly r2a) and FSCs but over time their most posterior descendants no longer produce ECs; most produce FSCs and FCs, while the most posterior produce only FCs. A subset of FC precursors accumulates beyond the interspersed somatic and germline cells of the developing germarium during the first two days after pupariation to form a newly-recognized structure that we termed an Extra-Germarial Crown (EGC) (see 48h APF in Fig. 14). These cells receive and then envelop the most mature germline cyst as it migrates out of the germarium in a remarkable morphological process of egg chamber formation that is quite different from egg chamber budding in adults. Soon after the formation of the first two egg chambers the morphology and marker gene expression pattern of the germarium resemble those of adult germaria, likely in response to establishment of a key niche signaling emanating from newlydifferentiated polar FCs in budded egg chambers.
EC, FSC and FC precursors might in principle adopt fate-appropriate AP locations during pupation stochastically or they might be guided by memory of an early-instructed fate, by selective adhesive interactions with germline cysts of differing maturity or in response to potentially longer-range extracellular signals. We found that the fate of developing precursors was drastically altered by genetic manipulation of Wnt and JAK-STAT signaling pathways. Altered Wnt signaling affected fate outcomes across the whole domain from anterior EC to FC precursors and throughout pupation, with higher Wnt signaling resulting in movement to more anterior locations. Loss of JAK-STAT pathway activity had a similar consequence to elevated Wnt pathway activity but the critical relevant times were not established. Elevated JAK-STAT activity greatly stimulated precursor proliferation, unlike the other manipulations of JAK-STAT and Wnt pathways, precluding analysis of any potential effects on precursor cell fates. Both Wnt and JAK-STAT pathways showed graded distributions, of opposite polarity and different times of inception, across the IC precursor region.

Identity, location and numbers of EC, FSC and FC precursors through pupal development
This study is the first to examine ovary development during pupal development through lineage analyses initiated at different times or by systematic analysis of fixed specimens and live imaging. Each approach included some challenges and leaves some issues unresolved but several key findings were conclusive and supported by all three approaches.
It is systematically challenging to conduct lineage analyses where only a single precursor is labeled in each developing ovariole because the number of dividing precursors that can potentially be labeled is large (more than 20 even at the earliest stages). This challenge was addressed by using very mild heat-shocks to initiate clones through hs-flp induction, a multicolor labeling technique that reduces the frequency of a specific color combination substantially relative to single-color clones, and by using information about clone frequencies and direct indications of composite clones to derive estimations of strictly single-cell lineage frequencies and yields. The results led to robust conclusions that many precursors at the start of pupation yield only ECs, a smaller proportion yield both ECs and FSCs (as well as FCs) and very few produce FSCs with no ECs. Common precursors for ECs and FSCs persist for at least the first 72h of pupation (later times were not tested), although they are exceeded in number by 72h APF by precursors that produce FSCs but not ECs. All such FSC lineages also included marked FCs in newly-eclosed adults. Thus, there is no cell or group of cells specifically assigned over this period to produce only FSCs or all FSCs.
The most distinctive group of precursors gives rise to FCs surrounding the most posterior egg chamber of newly-eclosed adults. These Fas3-positive cells, which we named Extra-Germarial Crown (EGC) cells in recognition of their location and appearance, accumulate from a handful to over twenty during the first 48h APF and then participate in an unprecedented set of dramatic morphological changes that produce the first egg chamber. During this process the most mature germline cyst leaves the germarium without accompanying ICs, enters EGC territory and moves a considerable distance from the germarium as it is enveloped by EGC cells. EGC cells were first defined in fixed images as expressing both TJ and Fas3 but live imaging showed that a few cells beyond the TJ-positive domain are also incorporated as FCs. Towards the end of the budding process cells posterior to the new egg chamber are intercalated to a width of two cells or less, forming a basal stalk, and a secondary EGC of Fas3-positive cells has accumulated between the germarium and the budded egg chamber. Secondary EGCs likely receive the second most mature cyst from the germarium but we have not observed that budding process directly to know whether the budding cyst retains germarial ICs or principally recruits secondary EGCs as FCs, as suggested by lineage analyses. The budding of subsequent egg chambers likely mirrors the process in adults as there is no longer a recognizable EGC and the two most mature germline cysts are now surrounded by Fas3-positive cells within the germarium.
Our conclusions about the origin of adult FSCs and those regarding derivation of the first FCs from a newly-recognized EGC structure have some common ground and some differences relative to another recent study that used single cell RNA sequencing to explore the diversity of precursors in mature larvae and their adult fates. Of most relevance here, it was suggested that a distinct group of cells characterized by bond-GAL4 expression, termed "FSCP"s for FSC and FC Progenitors, and located posterior to ICs, give rise to FSCs and FCs but not ECs (Slaidina et al., 2020). The principal supporting evidence is a cell lineage study with bond-GAL4 induced cell marking intended to initiate during larval stages of development (by shifting from 29C to 18C shortly after pupariation). Lineages were found in 2d-old adults to include germarial FCs, implying that an FSC had been labeled in newlyeclosed adults (FSCs were not scored directly) but labeled ECs were also seen. The authors proposed that marked FSCs arose from FSCPs and marked ECs arose from labeling a different cell type, presumably an IC prior to pupariation (due to broader expression of bond-GAL4 than bond). However, no clear evidence was presented that lineages had been initiated at the intended time or that ovarioles containing both marked FSCs and ECs contained two independent lineages; also, the frequency of ovarioles with marked FSCs (rather than FSCs or FCs) and no marked ECs was not reported because that can only be ascertained by scoring those cell types directly in newly-eclosed adults (Slaidina et al., 2020).
We repeated a bond-GAL4 lineage analysis. We found that the time of initiation for clones containing ECs and FSCs was not controlled effectively by temperature shifts (in our hands) but was nevertheless likely to be prior to pupariation in most cases. We saw aggregate outcomes broadly similar to those previously reported, with frequent labeling of ECs, FSCs and FCs, though we saw a much higher fraction of ovarioles containing only labeled ECs. Among 36 ovarioles with marked FSCs, all but one also included marked ECs. Also, because we examined newly-eclosed adults we could cleanly distinguish and score all FSCs and FCs produced during development and we found that marked FSCs were always accompanied by marked FCs, while some ovarioles contained marked FCs but no marked FSCs. The labeled ovarioles contained considerably more marked ECs and FSCs than in our MARCM analyses. By making a simplifying assumption (a guessed approximation of reality) that each ovariole contained exactly two marked lineages, we calculated the frequency and content of different types of lineage originating from single cells. Both the proportions of different types of lineage and the average number of marked ECs and FSCs within each lineage corresponded quite closely to the results obtained by inducing clones with hs-flp at pupariation. Those lineages, in both cases, included only ECs, ECs together with FSCs, but almost never FSCs without ECs. Thus, lineage analysis with bond-GAL4 was not as informative as inducing lineages with hs-flp (using MARCM and multicolor marking) because it did not prove possible to define the time of lineage initiation through temperature manipulation or to study single cell lineages directly but the results were nevertheless fully in accord with our observations of lineages induced using hs-flp.
Our MARCM studies showed that the "decision" of a precursor to become an FSC rather than an EC plays out over a long period of time and that the majority of FSC-producing precursors have not segregated from EC-producing precursors until mid-pupation. Hence, the population of 14-16 adult FSCs cannot possibly derive largely or entirely from a distinctive cell type that is an FSC/FC-specific precursor at the end of larval development, as has been suggested (Slaidina et al., 2020). Equally important, pupal precursors that produce FSCs almost invariably also give rise to FCs of newly-eclosed adults, showing that there is no strict delineation of FSC precursor fates from either EC or FC fates at the start of pupation.
The expression pattern of the bond gene was described as spanning all ICs and then refining in late third instar larvae to the extreme margin of ICs and slightly posterior (Slaidina et al., 2020). Our live imaging studies suggest that cells in the latter locations are likely to contribute substantially to the EGC and then to FCs of the first budded egg chamber, a location where bond-GAL4 activity was later detected in our study. Moreover, we found that bond-GAL4 induced lineages that labeled only FCs in the terminal egg chamber of newly-eclosed adults if animals were raised at 29C (to elevate bond-GAL4 activity and inactivate inhibitory GAL80 activity) for periods that included the first 2d or the last 3d of pupation. Although the window of thermosensitivity will need to be refined further, these results suggest that bond expression and bond-GAL4 activity may be selectively enriched in EGC cells and their products over a period preceding and following the budding of the first egg chamber at around 56h APF.

Extracellular signals that govern precursor movements and fate
We identified two signaling pathways that influence the AP location and consequent fate of precursors. The relevant tests all involved altering the genotype of a labeled lineage within the context of a large excess of normal precursors and therefore report consequences that are cell autonomous and reflect the behavior of labeled cells relative to normal neighbors. Reduction of Wnt signaling greatly reduced production of r1 ECs, suggesting that Wnt pathway activity must exceed a certain threshold for a cell to remain in the anterior third of the developing germarium. A Fz3-RFP reporter of Wnt pathway activity is uniformly highest in this region throughout pupation, with the absolute magnitude increasing significantly over the first 36h APF. The production of FSCs was increased by elimination of Wnt pathway activity (using a null arr mutation) but was unaffected by just reducing pathway activity (by expression of dominant-negative TCF). These observations suggest that in more posterior regions where Wnt pathway activity normally declines roughly linearly towards zero, posterior movement from an EC-producing to an FSC-producing location is restrained by even low pathway activity. The influences of Wnt pathway magnitude during pupal development are very similar to those observed on AP location and EC/FSC choices in adult ovaries (Reilein et al., 2017).
JAK-STAT signaling also affects pupal ovary development but our investigations are less complete than for Wnt signaling because we have not conducted a detailed time-course of lineage analyses and because of the nature of the initial phenotypes we identified. We found that increasing JAK-STAT activity greatly increased the number of labeled cells per clone and in almost all cases greatly distorted germarial morphology, preventing us from classifying and counting marked cell types.
Future tests will require a second genetic change to reduce precursor cell proliferation.
In normal developing ovaries we cannot clearly detect JAK-STAT activity at early stages.
Reporter activity was seen only after the first egg chamber has budded and was seen in a posterior to anterior gradient that intensified over time, suggesting that, as in adults, polar cells may be a critical source of this graded JAK-STAT signaling pattern. Because pathway activity is initially low and remains low in anterior ICs genetic overexpression of STAT is likely to increase pathway activity very substantially over normal levels in anterior precursors and in all precursors over the first 48h APF, consistent with the large effects we observed on cell proliferation. JAK-STAT pathway manipulations in either direction produced large effects on cell proliferation in adult germarial cells (Vied et al., 2012).
Loss of JAK-STAT pathway activity from the beginning of pupation onwards resulted in a large cell autonomous reduction in the production of FSCs and FCs, with a corresponding increase in EC production, closely resembling the effects of excess Wnt pathway activity. The temporal focus of this effect has not been defined but it seems most likely to be in the latter stages of pupation because it is only during that time that we can detect significant levels of pathway activity using a stat-GFP reporter.
Results from elevating Wnt pathway activity from -2d onwards show that two days can indeed suffice to reduce FSC production substantially in favor of ECs. Stat mutant precursors did not show any loss in cellular yields but very few of those cells and their descendants were likely in sufficiently posterior locations to be exposed to JAK-STAT ligand. We speculate that graded JAK-STAT pathway activity during the second half of pupation is important for stimulating proliferation of more posterior cells while anterior ICs slow and cease division.
Thus, our evidence to date suggests that graded Wnt signaling throughout pupation and graded JAK-STAT signaling during the second half of pupation are key organizers of the distribution of cells along the AP axis and their consequent final acquisition of EC and FSC fates. The relative proportions of adult ECs and FSCs produced are in part set by specifying a territory of precursors that give rise only to ECs and through the amplification of more posterior early precursors that produce both ECs and FSCs. These territories are influenced at the level of individual cells by the strength of Wnt pathway activity and a dependence on JAK-STAT signaling, suggesting that the evolving patterns of activity of these two pathways may be substantially responsible for dynamically defining EC, EC/FSC and FSCproducing domains. EC and FSC numbers also depend on the changing rates of precursor divisions in different locations. We speculate that JAK-STAT signaling has a key role in selectively promoting posterior cell proliferation after mid-pupation but we do not know which signals might regulate earlier proliferation, which also appears to be graded from anterior (low) to posterior.
It is likely that additional signaling pathways influence both the proliferation and location of precursors during pupal development. Delta signaling to Notch from terminal filament cells is important to specify adjacent cells as non-dividing Cap Cells roughly co-incident with the start of pupation (Gancz and Gilboa, 2013;Gancz et al., 2011;Panchal et al., 2017;Song et al., 2007;Yatsenko and Shcherbata, 2018). It was also previously shown that IC cells are distinguished from more posterior basal stalk cells prior to pupation through exposure to higher levels of Hh, emanating principally from terminal filament cells (Lai et al., 2017). It will be interesting to test whether Hh signaling affects the preference of a precursor to become an EGC or an IC, or EC, FSC and FC fates among ICs.

Coordination between germline and somatic ovarian cell development
Ovary development not only specifies suitable numbers of FSCs, GSCs and their niche cells but also prepares adults for producing mature eggs shortly after eclosion by producing a succession of maturing germline cysts in the germarium followed by four egg chambers. We have learned more about how pupal oogenesis is organized for both somatic and germline cells, providing a foundation for future exploration of coordination between the two.
It has been shown previously that germline differentiation is repressed globally prior to recruitment of anterior PGCs as future GSCs, at least in part through Ecdysone signaling acting on somatic cells (Gancz et al., 2011;Zhu and Xie, 2003). Then, late in third instar larvae, Ecdysone acts positively in somatic cells, via the Broad transcription factor, to initiate GSC differentiation, potentially in two steps of initial differentiation, evident from Bag-of-marbles (Bam) induction, and then cystocyte formation, revealed by the transition of the spectrosome to a fusome (Gancz and Gilboa, 2013;Gancz et al., 2011). Those studies did not, however, reveal whether Bam-GFP expression or PGC division initiates simultaneously in all PGCs distant from Cap Cells or in an anterior-posterior gradient at the larval /pupal transition or how germline differentiation proceeds thereafter. The differentiation of germline cysts in the adult germarium depends on interactions with ECs and this has also been seen in some cases during pupal development even though many EC precursors are still dividing during early pupation. Normal Wnt pathway activity is among those shown to be important for this function in ECs and pre-ECs (Hamada-Kawaguchi et al., 2014;Luo et al., 2015;Mottier-Pavie et al., 2016;Upadhyay et al., 2018;Waghmare and Page-McCaw, 2018;Wang et al., 2015;Wang and Page-McCaw, 2018). It is not known what drives ordered posterior movement of germline cysts in adult germaria but this also may depend on EC interactions (Banisch et al., 2017). In future it will be interesting to investigate whether pupal or adult ECs have graded EC adhesive properties or send graded differentiation signals according to their AP position, instructed for example by the strength of Wnt signaling.
A second major focus for germline and somatic cell interactions is the budding of an egg chamber. The process necessarily involves a germline cyst leaving the germarium and establishing a lasting interaction with dividing FCs sufficient to envelop it in a monolayer epithelium. In an adult germarium a stage 2b cyst is enveloped by Fas3-positive FCs, which then proliferate and progress towards an epithelial morphology for another 12h before a rounded stage 3 cyst buds from the germarium. Remarkably, the order of events is reversed for the first egg chamber to bud from a pupal ovary, with a rounded germline cyst emerging from the germarium before being enveloped over a period of a few hours by extra-germarial Fas3-positive EGC cells. The common factor is that germline cysts in both cases progress through somatic cell territory (ECs and FSCs, or ICs) with low or absent

Staging of pupae
Third-instar larvae were sorted to select females, transferred to a vial with food and checked every hour to mark the time each individual developed into a puparium (white, but immobile, with small anterior spiracles). This time is 0 hours APF (after puparium formation). We found that keeping animals in the light at night and in the dark during the day on a 12h/12h cycle, allowed more of them to pupate during the daytime. We used a programmable outlet timer (Nearpow) together with a manual LED soft white nightlight (Energizer cat. 37099) installed in a dark incubator at 25C.

Dissection and staining of pupal ovaries.
Pupae were removed from the vial wall by adding a drop of water and transferred with forceps into a well of a Corning PYREX glass spot plate (Corning, containing either PBS or 4% paraformaldehyde solution with care taken not to damage or pierce the posterior end of the pupa. If the dissection was done in fixative solution, it could take no longer than 10 min (ovaries were rocked in clean fixative for a minimum of 15 min, but for no longer than 30 min). Ovaries from 72h APF and older had fewer tears in them when dissected directly into fixative. Using a dissecting microscope, pupae were held against the bottom of the well using forceps and the posterior tip was cut off using Vannas Spring Scissors (2.5mm straight edge, Fine Science Tools, Foster City, CA, item number 15000-08) about a third of the way in from the posterior end. The anterior part of the pupa was removed to a discard well. The posterior end was searched for ovaries (clear, spherical, striated structures) by gently dislodging fat tissue surrounding the ovaries so as to not cloud the well with debris. This was done by gently tearing apart the contents of the pupal case or swirling them in the well using two pairs of forceps. As soon as they were spotted, ovaries were transferred to a new well of fixative by the following process. A 20µl pipette tip with a pipettor set to 20µl was coated in 10% Normal Goat Serum (NGS) by pipetting up and down several times. The pipettor was set to 5µl to pipette up the ovary from the dissection well to transfer into the fixative well. Ovaries were fixed by rocking ovaries in the well for 15 min at room temperature. Fixative was removed with a 1000µl pipette set to 270µl to slowly and carefully pipette up the liquid. The 1000 µl pipettor was used to remove liquid after all subsequent steps. Ovaries were rinsed 3x in 2% PBST 0.5% Tween solution for 5, 10, and 45 min at 4C by rocking very slowly in a horizontal plane. The glass dissection dish was covered with a large pipette tip box.
Ovaries were rinsed for 30-60 min in 10% NGS solution at 4C, and then rocked gently in primary antibody overnight. Ovaries were rinsed 3x with 0.5% PBST and then incubated for 1 hour (covered, cold room, rocker set on 2) in secondary antibody diluted 1:1000 with 0.5% Triton. Rinse ovaries 2x with 0.5% PBST and 1x with PBS. To mount ovaries, a 20µl pipette tip coated in 10% NGS solution as described above was set to 5µl and used to capture the ovary and transfer to a glass slide. 20µl of DAPI Fluoromount was added to a coverslip, and then placed on top of the slide with the ovaries. Care was taken to avoid pressing the coverslip and potentially disfiguring the mounted ovaries. For older pupal ovaries, specifically 72h APF and onwards, forceps were used to gently tear apart the ovarioles to separate germaria for clearer imaging. A sharpie was used to draw arrows to mark the location of the ovaries.

EdU labeling.
Ovaries were dissected directly into 15µM EdU in Schneider's insect medium, incubated for 1h at room temperature and then fixed for 10 min at room temperature in 4% paraformaldehyde. EdU incorporation was detected using the Click-iT Plus EdU Imaging Kit C10637 (Life Technologies). mounting medium for all other experiments. Images were collected with a Zeiss LSM700 or LSM800 laser scanning confocal microscope (Zeiss) using a 63x 1.4 N.A. lens.

Adult ovary fixation and staining.
Ovaries were fixed in 4% paraformaldehyde in PBS for 10 min at room temperature, rinsed three times in PBS with 0.1% Triton and 0.05% Tween-20 (PBST), and blocked in 10% normal goat serum (Jackson ImmunoResearch Laboratories) in PBST. Ovarioles were incubated in primary antibody for 45 min for multicolor lineage experiments and overnight for all other experiments. Ovarioles were rinsed in PBST three times and incubated for 1 h in secondary antibodies diluted 1:1000.

Imaging and quantitation of Frizzled3-RFP.
Fz3-RFP flies (made by Ramanuj DasGupta, Genome Institute of Singapore and a gift from Erika Bach, NYU, USA) were dissected and stained for Fas3 and DAPI. Ovaries were imaged with a 63 × 1.4 N.A lens on a Zeiss LSM 800 confocal microscope (Carl Zeiss). Zeiss Zen Blue was used to acquire microscope images of 1024x1024 pixels at 16-bit depth with line averaging 2. Zeiss Zen was used to measure the intensity of RFP in cells along an edge of a germarium in a line from just next to Cap cells and ending at the border of Fas3 expression. Measurements were taken in 5 µm distance increments within a z-stack range that did not exceed 8 µm. If there were more than one cell within the same 5 µm increment, both cells would be measured. If there were no suitable cell within the 8 um z-stack range at a specific 5 um increment, no measurement was collected. Fz3-RFP intensities were measured by circling the nucleus of a cell in the DAPI channel using the spline contour tool and obtaining the average intensity value in the 561 channel for the enclosed area from the Z-stack with the highest RFP intensity. The 561 laser intensity was set at 1.1% for 21-36h APF germaria and reduced to 0.3% beginning at 48h-APF through adult in order to keep RFP within the linear range. Because germaria in early z-slices were brighter than germaria in later z slices, RFP intensities for each germarium were normalized to the brightest cell on the edge of that gemarium. Normalized intensities were then averaged for each time point. In order to reflect the relative RFP brightness from 27h to 36h-APF, the averaged, normalized values were divided by the laser intensity at which they were taken. The averaged normalized values for every distance along the 27h-APF and 36-APF curves were divided by 1.1. The averaged normalized values for every distance along the 48h-APF, 60h-APF, and Adult curves were divided by 0.3. The 27h-APF, 36h-APF, 48h-APF, 60h-APF, and adult were then normalized to the highest value in all of the graphs to reflect relative intensities while preserving the Wnt gradient shape. In order to depict the relative RFP brightness for the 21h and 27h-APF germaria in the figure, output levels from 255 to 100 in Photoshop. The distances from Cap cells to the border of strong Fas3 expression on the edge of the germarium and to the posterior end of the germline in the center of the germarium were also measured for every germarium and averaged for each time-point.
Images were acquired every 15 min with 2.5 µm between Z-stacks. Clones were generated in the GFP/RFP ovaries by two 10 min heat shocks at 37° spaced 8 hours apart, 48 to 72 hours before imaging in order to generate GFP-only and RFP-only cells that were easier to track. Volocity Software (Quorum Technologies, Puslinch, Ontario, Canada) was used to make the 4D reconstruction of the UbiGFP labeled ovary. All nineteen z slices covering 45 µm were used in the reconstruction; however germaria drifted during imaging such that FCs covering approximately one quarter of the budded egg chamber on the lower germarium were not imaged. Cells were colored with Procreate software (Savage Interactive, North Hobart, Australia). Germline cells in the UbiGFP movie were identified based on their larger size and paler labeling with GFP compared to somatic cells.

Markers and clonal analysis of mutant genotypes.
Flies with alleles on an FRT42D or FRT82B chromosome were used in MARCM experiments: Two copies of stat-GFP were required to obtain a GFP signal in pupal ovaries (one is sufficient in adults).

Lineage analysis by MARCM
Pupae were heat shocked at 33C for 12 min at the number of days indicated before eclosion. Flies were dissected on the day of eclosion (0d MARCM) or collected on the day of eclosion and dissected 2 days later. Ovaries were stained for Vasa, Fas3 and GFP. Cells were scored as r1 or r2a based on measurements that were performed to determine the lengths of these regions. To determine the boundaries of each region we examined EdU incorporation in ovaries of newly eclosed flies and measured the distance from cap cells to the end of Region 1 as determined by the presence of cysts that incorporated EdU. Region 1 cysts are going through mitosis and Region 2a cysts have the complete complement of 16 germ cells. The length ratio of Region 1:2a is about 3:1 in newly-eclosed adults.

bondGal4/G trace experiments
bond-GAL4/G-TRACE flies (a gift of Ruth Lehmann, NYU medical center) were crossed at 18C and either kept at 18C, shifted after egg-laying to 29C and shifted back to 18C in mid-pupation (flies took 6 more days to eclose at 18C); or shifted to 29C 3 to 4 days before eclosion. Flies were dissected on the day of eclosion.

Inferring clone types, FSC and EC numbers in newly-eclosed adults from scoring 2d-old adults
In order to make quantitative conclusions about the developmental process up to eclosion from scoring clonal outcomes in 2d-old adults we made the following assumptions and adjustments. We assumed that EC numbers and locations did not change during this 2d interval because ECs generally appear to be quite long-lived. ECs are, however, produced from FSCs during adulthood, so it is likely that a small number of marked ECs scored at +2d were produced from FSCs after eclosion. The proportion of marked ECs arising from FSCs is very likely small based on known rates of FSC conversion to ECs and the observed high frequency of FSC clones with no ECs (44%) from -2d samples. Individual marked FSCs can be lost, mainly by becoming FCs, or amplified at high rates during adulthood. If one or more marked FSCs were present at 0d but became FCs by +2d those marked FCs would reside around the two germarial cysts or in the two youngest egg chambers because egg chambers bud roughly every 12h, allowing four cycles of FC recruitment over 2d. We therefore scored an ovariole as containing an FSC clone (FSC-only or FSC/EC, according to EC content) if there were any marked FCs up to and including the second egg chamber, even if there were no FSCs at +2d. In those cases we scored the number of marked FSCs as zero. On average, the number of marked FSCs should remain constant from 0d to +2d, so we should obtain a very good estimation of FSC numbers at 0d by scoring the numbers of marked FSCs at +2d, provided we include all examples where marked FSCs were lost altogether as containing zero FSCs. By using these guidelines we could surmise the exact number of different types of clone at 0d and the exact number of constituent ECs and FSCs from data collected by scoring clones in 2d-old adults.

A. Assuming independent recombination events
The proportion of marked ovarioles with a lineage originating from a single marked cell was first estimated by assuming that the probability of each dividing precursor to be labeled was independent.
The estimate derives principally from the observed frequency of labeled ovarioles but it is also dependent (to a lesser extent) on the number of potential target cells. Different numbers of assumed target cells are used in the calculations below to demonstrate that estimated single cell frequencies are not greatly altered by this value.
For the -6d to -2d time course, the average frequency of labeled ovarioles was about 60%. If there were only 6 target cells (estimated for -6d), assuming a binomial distribution (independent probabilities), the chance of any one cell giving a clone is p and the chance of 6 cells not giving a clone is (1-p) 6 . The observed no-clone fraction was 0.4, so (1-p)= 0.86 and p =0.14. Hence, the single clone frequency among 6 cells is 6p(1-p) 5 , and the frequency among labeled ovarioles is 1/0.6 fold higher (because 60% were labeled). Therefore, the expected frequency of single-cell lineages is 6x0.14x(0.86) 5 /0.

B. Assuming lower frequencies of single lineages based on the frequency of ovarioles with marked ECs and FCs
In MARCM and multicolor lineage experiments scored in newly-eclosed adults we observed a higher than expected frequency of ovarioles with labeled ECs and FCs but no FSCs, which we assume represent two distinct lineages because the marked cells are discontinuous. The observation indicated that single cell lineages frequencies were lower than calculated (above) assuming independent probabilities of precursor marking. Based on the EC/FC clone frequency and the proportion of unlabeled ovarioles we estimated that a specific proportion of labeled ovarioles harbored a single lineage, while the rest contained exactly two lineages. We settled on values that were roughly consistent with the observed frequency of ovarioles with marked ECs and FCs.

B1 MARCM clones induced 5d before eclosion
For the MARCM experiment examining newly-eclosed adults 5d after heat-shock, 67% of ovarioles were unlabeled. Single cell lineage frequencies and labeled cell content were calculated, as described below, assuming that 30% of ovarioles with marked cells derived from a single marked cell, while the remaining 70% derived from two marked cells. In all cases, EC/FSC and FSC-only categories include clones with and without FCs (aggregated).

B2 Multicolor clones induced 5d before eclosion
Analogous calculations for multicolor GFP-only clones where EC plus FC category was 8% and assuming 70% of marked ovarioles have lineages derived from one cell, and 30% have two lineages.    Also, by this time there are EC precursors that no longer divide (estimated at 5 for -4d and 18 for -3d from MARCM lineages examined in +2d adults, so perhaps 11.5 at -3.5d, to give a total of 45.4 precursors. So, estimated proportion of precursors that are pFCs is 13.3/45.4 = 29%.

B4 bond-GAL4/G-trace lineages
Single lineage frequencies and content for clones initiated by bond-GAL4were calculated in a slightly different way (considering ovarioles with only marked ECs to calculate EC-only single lineage and then ovarioles with labeled ECs and FCs clones to calculate the FC-only lineage frequency) because there were almost no ovarioles with only marked FCs.
Here we assumed that every labeled ovariole had exactly two lineages.          Time-stamped frames from a 3D reconstruction of a movie of 14h 45 min starting at 48h APF and showing budding of an egg chamber. All cells are labeled green from a ubi-GFP transgene. Three cells that were visible near the surface of a 3D reconstruction are colored in red, pink, and yellow in the lower germarium, together with three other cells (open circles) that were in interior z slices. All marked cells were tracked using Zeiss Zen software for 60 time points over 14h 45 min, except for the red cell which could not be reliably tracked after 10h and the yellow cell in the lower germarium, which was not tracked before 5h 45min. The posterior germline cyst in the lower germarium is highlighted in green.
Cyst cells were distinguishable from somatic because the nuclei were larger and paler green. The cyst slid past adjacent ICs, which ended up either on the cyst (red, pink cells from lower germarium) or (yellow cell in top germarium) anterior to the cyst in a secondary EGC (pink bracket). The cyan cell in the primary EGC (indicated by yellow brackets) adopted progressively more anterior positions on the germline cyst after the cyst moved into EGC territory. The white cell was originally in a location posterior to the EGC (and is expected, from fixed images, not to express TJ) but then moved into EGC territory by 3h 45min and ended up close to the cyan cell on the budded germline cyst. The location of the cyan and white cells covering the posterior of the cyst can be seen clearly in Fig. S3. The yellow cell in the upper germarium was initially located immediately anterior to the posterior cyst and moved into the secondary EGC (magenta brackets), as seen more clearly in Fig. S2A   Cartoon of the fate of single precursor types from 0h APF until 60h APF deduced from lineage studies and pupal ovary imaging. Most precursors (green) generate only adult ECs, with more anterior (left) precursors producing only r1 ECs (top row) and more posterior precursors also producing r2a ECs and more cells in total (second row). Filled green cells have ceased division. Most FSC-producing precursors also produce ECs. A few, more anterior precursors (third row) produce no FCs but most (fourth row) also produce FCs. Precursors that produce adult FSCs but no ECs also produce FCs (fifth row). Precursors at the posterior of developing germaria produce only FCs. A subset of their progeny accumulate posterior to the germarium from 24h APF onward in the EGC and most cells in the EGC will become FCs of the first budded egg chamber. More posterior precursors produce larger numbers of progeny because anterior cells divide more slowly and arrest division prior to eclosion.              Figure 14