While PSC formation is undoubtedly a dynamic process, current knowledge is derived exclusively from static timepoints using fixed embryos (Figure 1A). Thus, we sought to visualize the entirety of PSC migration in real time. To identify and track PSC cells live, we first confirmed their location at various time points in fixed embryos using the accepted markers, Antp and Odd, where co-expression demarcates PSC cells. We characterized PSC positioning in stages ranging from migration onset, stage 14, until migration completion, stage 16 (schematized in Figure 1B’–F’). Prior to migration, the PSC appeared as a compact cluster of cells (Figure 1B, green underline), it appeared more elongated through mid-migration (Figure 1C–E), and ultimately it assumed a compact, clustered organization at the posterior of the lymph gland (Figure 1F). During migration and upon its completion, the PSC flanked Antp + cardioblasts of the dorsal vessel (Figure 1F, brackets). Though not explored further here, ECM components, Perlecan and Viking, were expressed around the PSC collective (Figure 1—figure supplement 1A, B), including between the PSC and the adjacent cardioblasts (CBs) of the dorsal vessel.

The prospective cells of the LG and CBs lie just under a sheet of epidermis that itself undergoes dramatic movement in late-stage embryos, concurrent with LG and CB migration. Leading edge (LE) epidermal cells move to the dorsal midline during dorsal closure, which is complete by stage 16. It has previously been shown that CBs migrate independently of LE epidermal cells (Haack et al., 2014; Balaghi et al., 2023). To address whether the LG/PSC also migrates independently, we assessed LG positioning in stage 17 mutants in which dorsal closure had stalled. Both heterozygous and homozygous hlh54f⁵⁹⁸ mutants (Figure 1—figure supplement 2A, B, C, respectively) exhibited defects in closure, evident by epidermal leading edge (LE) cells that stalled at lateral positions (LE indicated by cyan lines). Embryos stained with p-Tyr to label cell membranes and Odd to label the LG were scored for position of the LGs relative to the stalled LE. In over two-thirds of cases the LG had migrated past the stalled LE (Figure 1—figure supplement 2D), indicating that LG migration is generally uncoupled from this large-scale epidermal movement, similar to the case for CBs.

To our knowledge, a sufficiently bright, stable marker that labels only PSC cells during migration stages does not exist. Thus, for live-imaging we combined Hand-RFP, which marks lymph gland (including the PSC) and CBs, with a myristoylated GFP driven by Antp-GAL4 (Antp >myrGFP) to mark the PSC but not other lymph gland cells (Figure 1G–J). We performed timelapse imaging on embryos beginning after PSC specification (stage 13) until the PSC completed migration and morphogenesis (stage 17). We observed that the PSC moved as a collective for the entire migration (Figure 1G; Figure 1—video 1). During migration many PSC cells extended protrusions—some reached far anteriorly, encasing the lymph gland (Figure 1H’, bracket) while others were shorter or branched (Figure 1H’’’, arrowheads). The protrusions exhibited no apparent directional bias but were more common at non-dorsal vessel-adjacent surfaces of the collective. During migration some cells shifted positioning with respect to one another (Figure 1—video 2). For example, two lateral PSC cells that abut early on (Figure 1I, 0m, arrowhead) become separated by interdigitating cells (Figure 1I, 8-40m, magenta and green dots). In general, however, most cells remained at relatively fixed positions within the collective—particularly the posterior-most, dorsal vessel-adjacent PSC cells (Figure 1I, orange dots and Figure 1G, arrowheads).

Overall, the features observed during live-imaging lead us to characterize PSC formation as collective cell migration. The intricate, dynamic extensions we observed on PSC cells reflect an active actin-based cytoskeleton—a common theme amongst migrating cell collectives, where these extensions are often used for sensing environmental guidance cues. Most importantly, even as some PSC cells shifted their position relative to one another, the group accomplished directional movement while maintaining coalescence. The persistence of PSC-to-PSC cell contacts despite internal movements suggests coordination within the group to orchestrate remodeling of cell adhesion such that a unified directional movement is achieved. Altogether, these features are consistent with this as an example of collective cell migration, and suggest the importance of collectivity in building the PSC.

Crucially, live-imaging also uncovered that the PSC was within signaling distance of various muscles for the entirety of its migration. It is well-documented that muscles can supply positional information to nearby tissues (Anllo and DiNardo, 2022; Scimone et al., 2017; Witchley et al., 2013), and our imaging indicated that visceral muscle (or, visceral mesoderm; Vm) and CBs were each near the PSC. We live-imaged with a dorsolateral vantage point which revealed that the PSC migrated dorsally in synchrony with CBs and Vm (Figure 1J; Figure 1—video 4). The PSC remained laterally affixed to Antp +CBs (Figure 1J’, blue) and slightly dorsal to, but not contacting, Antp +Vm (Figure 1J’, orange) throughout migration.

Based on segmental positioning, we suspected the PSC was also near alary muscle (AM)—segmentally repeating muscles that attach internal organs to the body wall. We live-imaged PSC and AM migration from a dorsal vantage point with Antp >CD8:GFP and org-1-HN39::RFP, a marker with expression restricted to AMs (Figure 1—video 5). Indeed, the first AM was immediately lateral to the PSC for the entirety of their synchronized migration to the dorsal midline (Figure 1K, PSCs outlined in green). We confirmed this proximity by fixing org-1-HN39::RFP embryos and staining for Antp, which revealed that at late stages, the A1 AM ensheathed the PSC (Figure 1—figure supplement 1D), with some fibers encasing individual PSC cells (Figure 1—figure supplement 1E).

Thus, live-imaging provided invaluable insights, advancing our knowledge of PSC formation from that attainable through analysis of fixed preparations (Figure 1A), by revealing the mode of PSC migration and the spatiotemporal dynamics of the PSC relative to nearby tissues (Figure 1L). The sustained proximity of alary muscles, visceral mesoderm, and cardioblasts to the PSC made each a strong candidate for influencing PSC positioning.

Having identified multiple candidates, we next investigated whether any or all of these muscles were necessary for proper PSC positioning. To test a role for AM, we ablated it by expressing the pro-apoptotic factor, grim, using the AM-specific AME_r-GAL4. Ablation was successful, evidenced by lack of GFP-labeled AMs (Figure 2B) compared to controls (Figure 2A). Based on our characterization of the wildtype PSC (Figure 1), we developed criteria to score PSCs at the culmination of their migration as positioned ‘normally’ or ‘abnormally’. A ‘normal’ PSC must be (1) coalesced within one nuclear diameter of one another, (2) adjacent to the dorsal vessel, and (3) at the same dorsal-ventral position as the posterior-most progenitors of the lymph gland. Surprisingly, AM ablation had no impact on PSC positioning (Figure 2C) nor the total number of PSC cells (Figure 2—figure supplement 1), indicating that AMs are not required for proper PSC formation.

To determine whether Vm was involved in forming the PSC, we examined biniou mutants in which the Vm is genetically ablated (Zaffran et al., 2001). In controls, the Vm is apparent at stage 12 as columnar, Fas3+ cells (Figure 2D, brackets). By stage 16, Vm has surrounded the gut and constricted it into sections (Figure 2E, arrow is first Vm constriction that separates the first two gut sections). By contrast, in bin mutants, minimal Fas3+ Vm was detected at stage 12 (Figure 2G), and it was absent at stage 16 (Figure 2H). The majority of bin^R22 mutants and bin^R22 /binS^S4 transheterozygous mutants had abnormal PSC positioning (Figure 2I, arrowheads) compared to heterozygote controls with normal PSC positioning (Figure 2F, outlines; quantified in 2 J). We detected no Bin protein in PSC cells (Figure 2—figure supplement 1B), indicating PSC positioning defects in bin mutants did not originate PSC-intrinsically but rather were caused by lack of Vm.

To confirm a role for Vm in PSC formation, we ablated Vm by expressing the proapoptotic gene, hid, using a Vm-specific driver, bap-GAL4. Stage 11 sibling controls displayed characteristic organization of Vm precursors (Figure 2K): segmentally repeating mounds of fusion competent myoblasts (FCMs; between green and magenta lines) atop columnar founder cells (FCs; below magenta line). Stage 11 ablated embryos were largely missing FCMs (Figure 2N, brackets). Although most FCs were present at this stage, they exhibited substantially reduced Bin expression compared to controls (Figure 2—figure supplement 1C and D, quantified in 1E), suggesting improper differentiation. By stage 16 no Vm was detectable (Figure 2O). Vm-ablated embryos had abnormal PSC positioning (Figure 2P, arrowheads) more often than sibling controls (Figure 2M, quantified in 2Q). Taken together, these analyses definitively establish that an extrinsic cue(s) provided by Vm governs PSC positioning.

To address when the cue might be delivered, we analyzed jelly belly mutants and bagpipe hypomorphs, bap²⁰⁸. In both backgrounds, Vm precursors are present initially at stage 11 but do not differentiate or migrate properly from stages 13–16 (Figure 2—figure supplement 1 and Englund et al., 2003; Lee et al., 2003; Weiss et al., 2001). In both mutants, PSC positioning was unaffected (Figure 2—figure supplement 1K, N, and P; quantified in L and Q), revealing that early signaling from Vm is required for PSC positioning.

Finally, to assess whether cardioblasts of the dorsal vessel impact PSC positioning, we expressed hid or grim with the CB-specific driver, tinCΔ4-GAL4. We used conditions that restricted defects to CBs and left Vm unaffected (see Materials and methods). Whereas controls had uninterrupted strings of Mef2-labeled CBs (Figure 2R), the manipulated embryos had significant dorsal vessel ablation, evident by gaps in CBs (Figure 2S, bracket). We elected to analyze stage 14 and 15 embryos because these presented with appreciable ablation over larger distances compared to stage 16 onwards. Oftentimes both sides of the bilaterally symmetric dorsal vessel were ablated; sometimes only one side was ablated. For a given instance of dorsal vessel ablation, we documented whether the ipsilateral PSC was positioned normally. While the PSC occasionally appeared mis-positioned in non-ablated controls, for ablated dorsal vessel the ipsilateral PSC was mis-positioned more frequently (Figure 2T). These data suggest that dorsal vessel CBs influence PSC formation.

To identify a cue necessary for positioning the PSC, we sought a signal that is expressed in early Vm and/or CBs. The secreted glycoprotein Slit is known to be expressed in both tissues and has been implicated as a positional cue in multiple developmental contexts (Anllo and DiNardo, 2022; Kidd et al., 1999; Kolesnikov and Beckendorf, 2005; MacMullin and Jacobs, 2006; Rothberg et al., 1990). We confirmed that Slit is detectable in both Vm and CBs at relatively early stages of PSC migration (Figure 3A and B). Importantly, sli mutants exhibited mis-positioned PSCs more frequently compared to controls (Figure 3C and D, quantified in 3E), indicating a requirement for Slit in properly positioning the PSC.

To confirm that PSC cells can respond to Slit ligand, we examined the expression of the three Robo class receptors. We found evidence for only Robo1 and Robo2 expression in the PSC (Figure 3—figure supplement 1A, B). Thus, to test the requirement for Robo signaling, we analyzed single and double robo1 and robo2 mutants. For both single mutants, about 50% of embryos had abnormal PSC positioning (Figure 3I and J, arrowheads). About 75% of double mutants had abnormal PSCs (Figure 3K), a frequency which matches that of sli mutants. These results demonstrate that Slit signaling through Robo1 and Robo2 is required for properly positioning the PSC.

To test whether CBs were a source of Slit important for PSC positioning, we used tinCΔ4-GAL4 to prevent either production or release of Slit specifically from dorsal vessel using RNAi or sequestration via overexpression of the Robo1 receptor, respectively. We confirmed that two independent Slit RNAi’s eliminated nearly all Slit accumulation at CBs (Figure 3O’ , and Q’), while Robo overexpression resulted in more pronounced Slit accumulation on CB membranes (Figure 3S) compared to control CBs (Figure 3M’). PSCs were abnormally positioned more frequently under all challenge conditions (Figure 3N, P and R) compared to controls (Figure 3S). Thus, Slit produced from CBs is essential to properly position the PSC.

The additional requirement for Vm in PSC positioning and the presence of Slit expression in Vm suggested that Vm-derived Slit might also regulate PSC formation. To block Slit produced by Vm from signaling to PSCs, we sequestered Slit specifically in Vm using bap-GAL4 to overexpress Robo1. This manipulation prevents diffusion of Slit away from the tissue, as evidenced by distinct accumulation of Slit on Vm membranes (Figure 3Z) compared to diffuse Slit puncta in control Vm (Figure 3W). Under this condition, we indeed found that PSCs were mis-positioned significantly more often (Figure 3AA; quantified in 3BB). Taken together, these data suggest that Vm and CB-derived Slit signaling are both required for PSC positioning.

While the data thus far supported a simple model wherein Slit secreted from Vm and CBs acts directly on Robo receptors expressed by PSC cells to influence PSC positioning, further investigation revealed a more complex situation. Autocrine Slit-Robo signaling among dorsal vessel cells is important for proper polarity and organization of the vessel (Medioni et al., 2008; Qian et al., 2005; Santiago-Martínez et al., 2008). Intriguingly, in experiments where Slit was sequestered in Vm, we noticed defects in the organization of the cardioblasts (Figure 3AA, brackets; compare to control, 3 X brackets). This suggested that signaling from Vm was required for organizing the dorsal vessel—a possibility that had not been previously explored. To test this, we examined the dorsal vessels of bin mutants which lack Vm (Figure 2H). We identified multiple phenotypes, two of which we termed ‘twisted’ and ‘sunken’; the dorsal vessel in Figure 4B exhibits both phenotypes. We scored a dorsal vessel as ‘twisted’ (Figure 4B, bracket) when there was an apparent kink in the vessel such that the left or right side sat directly atop the other; thus, when viewing a single z-slice, a span of contralateral CBs appear to be missing since they are displaced above or below that focal plane. We defined a sunken dorsal vessel (Figure 4B’ and B’’) as those cases where the whole dorsal vessel, or part of it, was displaced ventrally, deeper into the embryo. Such a case is evident in the Figure 4B mutant dorsal vessel, which spans 8 µm of depth, compared to the sibling control dorsal vessel, which spans only 4 µm. Both phenotypes were significantly more frequent in bin mutants compared to sibling controls (Figure 4C). Together, the dorsal vessel defects observed upon removal of the Vm or from trapping Slit on Vm cells reveal a previously unrecognized role for Vm in dorsal vessel formation.

Our results indicated that lack of Vm causes abnormal dorsal vessel positioning, and that the dorsal vessel is required for PSC positioning. In addition, our live-imaging revealed that some PSC cells are affixed to CBs of the dorsal vessel (Figure 1). Thus, it became important to test whether all PSC phenotypes might be ‘passive’, explained by PSC attachment to a malforming dorsal vessel. Alternatively, the PSC defects could reflect a requirement for Robo activation directly in PSC cells. Without a PSC-specific driver to directly test the latter, we re-examined robo1,robo2 double mutants, this time including an additional marker that allowed us to observe PSC and CB defects independently and score for correlation between the two. If defective PSC positioning mostly correlated with defective CB positioning this would suggest a passive effect on PSCs by CBs of the dorsal vessel (Figure 4D, middle schematic). By contrast, if PSC cells were mis-positioned without a similarly mis-positioned CB nearby, this would suggest a requirement for intrinsic activation of Robo in PSC cells (Figure 4D, rightmost schematic).

The CB marker, svp-lacZ, labels two CBs in each hemisegment (Figure 4E, two hemisegments shown; Lo and Frasch, 2001). Normally, the PSC is compact and located adjacent to the compact Antp +CBs (Figure 4E, PSCs outlined). Svp co-labels the last of the four strongly Antp +CBs, and it labels the immediately posterior Antp- CB (Figure 4E, brackets). At the end of migration, Svp-LacZ +CBs, like all CBs, are well-aligned with their bilaterally symmetric counterparts. Thus, including Svp-LacZ in our analysis served as a useful registration marker that afforded the ability to detect bilateral matching, as well as shifting of CBs or the PSC along the anterior-posterior axis (Figure 4E–G, arrows indicate second set of Svp +CBs).

In robo1,robo2 double mutants, we found that mis-positioning of CBs and the ipsilateral PSC occurred more frequently compared to double heterozygotes (Figure 4H). We then further analyzed those instances of mis-positioning for whether the position of PSCs correlated with position of CBs. In about two-thirds of cases mis-positioning was correlated (Figure 4F and G, green arrowheads, quantified in 4I), suggesting that proper PSC positioning relies in part on proper CB positioning. However, in about one-third of cases PSC mis-positioning appeared independent of CB mis-positioning (Figure 4F and G, magenta arrowheads). We also observed PSC mis-positioning (Figure 4G, magenta arrowhead) when ipsilateral CBs were positioned normally (Figure 4G, brackets). These instances of non-correlation strongly suggest that Robo activation is required in PSC cells for their proper positioning.

Finally, we live-imaged slit mutants with Hand-RFP, Antp >myrGFP to examine the dynamics of PSC migration when signaling was compromised (Figure 5—video 1; Figure 5—video 2). In contrast to the robust collectivity noted throughout the migration of control PSCs (Figure 1G), this imaging revealed progressive deterioration of PSC integrity during migration. For instance, throughout imaging (Figure 5—video 1), one PSC cell (Figure 5A, cyan dot) bridged the gap between the Antp +CBs (Figure 5A, yellow brackets, 0 min) and posteriorly displaced CBs. Ultimately this PSC cell made aberrant contact with a CB from the contralateral side (Figure 5A, yellow dot, 70 min). This region had a persistent gap, possibly reflecting improper sealing of the dorsal vessel (Figure 5A, arrows, 112 min), as well as improper A-P alignment of Antp +CBs with their contralateral counterparts (Figure 5A, offset brackets, 112 min). Another PSC cell (Figure 5A, cyan arrowhead, 0–35 min) extended long protrusions along the peripheral edge of the PSC and was cleared by a macrophage (Figure 5A, yellow arrowhead, 42–49 min). At imaging onset in the same embryo (Figure 5—video 2), 2 PSC cells were laterally displaced (Figure 5B, arrowheads, 0–28 min) but connected to the main PSC collective by a thin protrusion (Figure 5B, green arrow, 0 min). These two cells detached from the collective (7–28 min), underwent cell shape changes (elongated at 42–56 min, brackets; highly protrusive at 77 min), and remained relatively stationary as the rest of the PSC and CBs migrated away (out of this focal plane).

Altogether, these data demonstrate that proper PSC formation requires (1) Slit signaling from the Vm and dorsal vessel where signaling is necessary for proper alignment and organization of the dorsal vessel, which, in turn, affects positioning of attached PSC cells; and (2) Robo activation in PSC cells for their persistent association with one another as a collective (Figure 6).

Prior knowledge of the embryonic PSC has been focused on PSC cell specification (Crozatier et al., 2004; Mandal et al., 2007). Thus, our live-imaging provides a substantial advancement by revealing PSC dynamics after specification, during migration. PSC formation is an example of collective cell migration, reminiscent of border cell migration in the Drosophila egg chamber and migration of the lateral line primordium in Danio rerio. These cell collectives respond to guidance cues, and their constituent cells are highly protrusive and shift position within the collective (Cliffe et al., 2017; Dalle Nogare et al., 2020; Haas and Gilmour, 2006; Peercy and Starz-Gaiano, 2020). PSC migration is no different (Figure 1)– we observe extension and retraction of protrusions on all surfaces of the collective except those in contact with CBs. This suggests that the migrating PSC explores and responds to its environment. Some cells of the PSC maintain their position adjacent to the dorsal vessel throughout the entirety of migration, while other cells exhibit fluidity within the collective. The positional shifts suggest remodeling of adhesive contacts between cells—albeit in a coordinated manner to achieve unidirectional migration for the entire cohort. Interestingly, circulating hemocytes can be seen interacting with the PSC—hemocytes are known to deposit and remodel ECM (Bunt et al., 2010), which could serve as a scaffold for adhesion molecules and as a substrate during migration. Indeed, we identified ECM components surrounding the PSC, including its interface with the dorsal vessel. The characteristics noted above could be further explored to improve our understanding not only of how the PSC is built, but more broadly of collective cell migration.

For improved live-imaging, we sought a PSC-specific marker but found most candidates to be expressed in only subsets of PSC cells (unpublished data); the lack of homogenous expression suggests PSC heterogeneity. The existence of fixed and fluid PSC cell positionings in the collective further hints at heterogeneity—perhaps there are different types and polarities of adhesive molecules. If PSC heterogeneity exists at this early stage, its purpose is unknown. One possibility is that the fixed, dorsal vessel-adjacent PSC cells act as ‘leaders’ while the more fluid PSC cells are ‘followers’, a common phenomenon in collective cell migration (Qin et al., 2021). The dorsal vessel has recently been shown to serve as a second hematopoietic niche component (Destalminil-Letourneau et al., 2021; Tian et al., 2023), so another possibility is that the fixed, dorsal vessel-adjacent PSC cells have a special function as intermediaries that coordinate communication between the niche components. That said, whether the niche components interact remains unknown. Finally, PSC heterogeneity in this early context could reflect a division of labor later, when the PSC regulates larval hematopoietic progenitors. Gene expression heterogeneity has been noted in the testis niche (Le Bras and Van Doren, 2006; Raz et al., 2023; Zheng et al., 2011), suggesting that this may be a conserved feature of niches which warrants further investigation.

We reveal that positioning the PSC requires input from different extrinsic sources. It is perhaps notable that the few well-studied examples of niche formation all involve extrinsic inputs. The mammalian intestine is organized into protrusive villi and invaginated crypts; Wnt and EphB3 signals emitted from the crypt base recruit niche cells from the crypt-villus interface to their final position in the compact troughs of the crypts (Batlle et al., 2002; Holmberg et al., 2006; van Es et al., 2005). The C. elegans gonadal niche cell relies on adjacent germ cell proliferation for propulsion towards its steady-state position at the apex of the gonad (Agarwal et al., 2022). A microniche of the mammalian hair follicle, the dermal papilla, originates via extrinsic Fgf20 signaling to dermal fibroblasts. The fibroblasts are recruited into a condensate via directed migration, and then the condensate is segregated deeper into the skin via epidermal invagination driven by reciprocal signaling between dermis and epidermis (Biggs et al., 2018). The Drosophila testis niche cells require FGF and Slit signals from Vm to migrate through the testis and assemble as a cap that is further compacted via actomyosin contractility (Anllo and DiNardo, 2022; Warder et al., 2024). Together with our findings on PSC formation, a paradigm for niche formation is emerging that, subsequent to niche cell specification, extrinsic input segregates niche cells from other constituents, and positions the collective toward one end of the tissue. Oftentimes positioning is accompanied by or precedes compaction of the recruited cells into a final niche shape/architecture.

We show that extrinsic Slit-Robo signaling is integral in establishing PSC position. Removing Slit causes PSC defects at the same frequency as removal of both canonical Robo receptors, indicating that in PSC formation, Slit only signals through Robo, and not through the non-canonical receptor, Dscam1. The PSC positioning phenotype was about 75% penetrant, and the severity of the defects varied. These observations argue that PSC formation relies on an additional cue(s) yet to be discovered. Evidence for an additional PSC coalescence cue exists in later larval stages, as manipulated Insulin Receptor signaling was shown to disrupt PSC coalescence (Tokusumi et al., 2015). Unfortunately, most pathways are difficult to test for a role in PSC formation due to lack of an embryonic PSC-specific tool with the necessary temporal control.

In the nervous system, one copy of the normal (WT) allele of Slit or Robo are typically sufficient for proper development (Kidd et al., 1999; Rothberg et al., 1988). In contrast, we find that about half of robo2,robo1 double heterozygotes or of slit heterozygote embryos have abnormal PSCs, suggesting haploinsufficiency in this process. Our observation in embryonic stages is supported by previous findings at later larval stages where both slit or robo2 heterozygotes have dispersed PSCs (Morin-Poulard et al., 2016). Perhaps haploinsufficiency in the PSC reflects different regulation and function of the Slit-Robo pathway from the nerve cord. Alternatively, because Slit-Robo signaling is required within both CBs and PSC for their proper development, the haploinsufficiency may reflect a combinatorial defect caused by diminished signaling in both tissues. Furthermore, we are surprised that robo2 and robo1 single mutants have a similar frequency of abnormal PSCs because in the embryo Robo1 is expressed in all PSC cells, whereas Robo2 appears to be expressed in only one. An intriguing possibility is that the Robo2-expressing PSC cell is the ‘leader’ in PSC migration, and without Robo2 signaling in the leader, the collective fails to migrate appropriately, or the followers fail to remain properly adherent. However, it may be that all PSC cells express Robo2, and our finding reflects limited detection ability with the Robo2-GFP reporter.

Morin-Poulard et. al. show that all larval PSC cells express Robo2, and it has a more prominent role than Robo1 in maintaining PSC coalescence (termed ‘clustering’ in their work) during larval stages (Morin-Poulard et al., 2016). In fact, they found Robo1 knockdown alone was insufficient to disrupt clustering. Taken together with our work, perhaps this indicates that Robo1 is more important for first establishing PSC coalescence and Robo2 is more important for maintaining it. Most manipulations occurred after the PSC had already formed, and therefore the changes reported cannot stem from defects during establishment but rather from dispersion of the cluster after it formed. One experiment without temperature control that could have diminished Robo signaling in PSC and CBs early (Antp >Robo RNAi), at about the time of PSC specification, generated a more dispersed late larval PSC compared to Robo knockdown after PSC formation. One can imagine that aberrant embryonic PSC positioning would become exacerbated and increasingly catastrophic as the PSC and nearby progenitors proliferate throughout larval stages.

Determining how Slit affects a given process is complex because it can act as an attractive or repulsive cue, or even affect cell adhesion, and in some instances the outcome relies on whether Slit is cleaved by Tok (Englund et al., 2002; Kellermeyer et al., 2020; Kolesnikov and Beckendorf, 2005; Kramer et al., 2001). Thus, elucidating the role of the Slit-Robo pathway in PSC formation is difficult. The existence and relative positioning of two Slit sources, the reliance on one for positioning of the other, and the inconsistent direction of PSC mis-positioning suggest that it is too simplistic to ask whether Slit is acting as a repulsive or attractive cue for PSC cells. Notably, mis-positioned PSC cells adopted aberrant contacts with CBs, LG, and pericardial cells of the dorsal vessel. When reported in the dorsal vessel, this inappropriate mixing of cells was attributed to improper polarity of cell adhesion molecules (Qian et al., 2005; Santiago-Martínez et al., 2006). Furthermore, Slit is important for maintaining adhesion amongst larval PSC cells (Morin-Poulard et al., 2016). Taken together with our data, we reason that during PSC migration Slit must affect the ability of PSC cells to properly adhere to the dorsal vessel and to one another. In this manner, Slit facilitates PSC cell association as a compact cluster.

The phenotypic variability we observe upon compromising Slit-Robo signaling indicates that downstream regulation is likely complex and dependent on the specific context, in time and space. Morin-Poulard et al., 2016 show that the Slit-Robo pathway maintains clustering of larval PSC cells via DE-Cadherin and Cdc42; however, our data indicate different effectors are at play during PSC formation, as we detect no DE-Cad in the embryonic PSC (data not shown). Additionally, whereas constitutively active Cdc42 in the larval PSC caused dispersion (Morin-Poulard et al., 2016), our attempt to elicit PSC dispersion with this manipulation in the embryo yielded normally coalesced PSCs (data not shown). Therefore, we reason that different downstream effectors are engaged by the Slit-Robo pathway for initial positioning of the PSC.

This study was limited by the lack of a truly PSC-specific driver—a tool that would improve live-imaging and allow testing of potential Slit-Robo pathway effectors. We and others observe Fasciclin III, a homotypic cell adhesion molecule, in the PSC (unpublished data and Mandal et al., 2007), however, it is first detectable too late in embryogenesis to function during PSC formation. Thus, the particular adhesion molecules that maintain association of PSC cells as they migrate remain unknown. Elucidating this tool will facilitate interrogation of how Slit-Robo signaling impacts PSC adhesion to the dorsal vessel and amongst itself.

Due to the necessity of autocrine Slit-Robo signaling in dorsal vessel formation (Medioni et al., 2008; Qian et al., 2005; Santiago-Martínez et al., 2006; Santiago-Martínez et al., 2008), it was challenging to distinguish whether effects on the PSC were secondary to mis-positioned CBs, or primary consequences of diminished Robo activation in PSC cells. Our analysis of robo2,robo1 double mutants along with the registration marker, svp-lacZ, strongly suggests both passive mis-positioning of PSC cells via attachment to mis-positioned CBs and a requirement for Robo activation directly in PSC cells. A direct requirement for Robo in PSC cells is further suggested in slit mutant live-imaging. Here, the lateral-most PSC cells separated from the main cluster even though no CBs were nearby (Figure 5B). Perhaps during normal migration these more ‘fluid’, lateral PSC cells possess a stronger requirement for Robo activation and require direct input to achieve that. In this view, those PSC cells that are passively mis-positioned in mutants are the same PSC cells that remain affixed to the dorsal vessel during normal migration. Alternatively, or additionally, passive mis-positioning of CB-adjacent PSC cells might occur due to improper polarization of adhesive molecules, whereas separation of lateral PSC cells from the collective might occur due to a total loss of adhesive molecules. Finally, it is possible that PSC cells do not intrinsically require Robo activation, but rather CB-independent PSC mis-positioning in sli or robo mutants could be a secondary defect caused by compromised Slit-Robo signaling in some other tissue. A PSC-specific driver to knockdown Robo intrinsically would be needed to test definitively the requirement for Robo in PSC cells.

The coalesced nature and posterior positioning of the PSC are its most prominent features. The consistency of these features suggested that they are under tight regulation, as we report here, and strongly indicates their relevance to function of the PSC. Although we have yet linked the precise architecture of the embryonic PSC to function, the specific architectures of multiple other niches have been shown to be functionally relevant. In the Drosophila gonad, a coalesced, compact niche is necessary for proper germline stem cell maintenance signaling, and for orienting stem cell divisions (Anllo and DiNardo, 2022; Warder et al., 2024). In the mammalian hair follicle, niche position determines stem cell fate (Rompolas et al., 2013). Drosophila neural stem cells are supported by glia, and the precise morphology of this niche is required for homeostasis of the entire nerve cord (Spéder and Brand, 2018). Most pertinently, recent work has shown that the larval PSC communicates amongst itself via a calcium signaling network. The integrity of this network is gap junction-dependent and is required for the PSC to emit proper levels of a progenitor maintenance signal (Ho et al., 2021; Ho et al., 2023). We find it reasonable to postulate that a dispersed PSC would exhibit defective calcium signaling such that the PSC would lack coordinated maintenance of progenitors, thereby leading to an imbalance in the ratio of progenitors to differentiated hemocytes in the gland. Future experiments will determine how PSC coalescence is functionally relevant to its regulation of hematopoietic progenitors.

Detailed information on the Drosophila strains used in this study is in Appendix 1—key resources table. Controls were GAL4 only, a cross to w¹¹¹⁸, or a sibling control identified by a fluorescent balancer. This study generated the following embryo genotypes by combination into a stock (available upon email request) or obtained through a cross:

Unless otherwise noted, embryos were collected on apple juice agar plates overnight at 25°, unless an experiment involved a Gal4, in which case collections occurred at 29°. The next morning embryos were dechorionated in 50% bleach.

Using a paintbrush, embryos were transferred from a collection basket to a 50/50 mixture by volume of heptane and 4% paraformaldehyde in Buffer B (16.7 mM KPO₄, pH 6.8; 75 mM KCl; 25 mM NaCl; 3.3 mM MgCl₂)(de Cuevas and Spradling, 1998). Embryos were fixed on a rocker for 15 min, then fixative was removed. An equal volume of MeOH was added and the vial was shaken vigorously for about 5 s to remove vitelline membranes. The heptane and MeOH mixture was discarded, and the embryos were rinsed three times with MeOH.

Embryos were staged according to Atlas of Drosophila Development (Hartenstein, 1993). Embryos were genotyped according to presence of a balancer chromosome: CyO, P({Dfd-EYFP}); or CyO,P({Wg-lacZ}); or CyO, hindgut-lacZ; or TM6, P{Dfd-EYFP}, Sb, Hu, e.

Manipulations were room temperature unless otherwise noted. Embryos were rehydrated in 50%MeOH/50% PBS (10 mM Na₂ HPO₄; 1.8 mM KH₂PO₄; 2.7 mM KCl; 137 mM NaCl; pH 7.4), followed by 100% PBS. Embryos rocked for 5 min in PBS with 0.1% Triton X-100 (PBST), then 1 hr in 4% normal donkey serum in PBST. Embryos were then transferred to a rocker at 4° until the end of the day or began incubation in primary antibody solution. The next day embryos were rinsed three times in PBST then rocked in PBST for 1 hr, incubated in secondary antibody solution for 1 hr, rinsed three times in PBST, rocked in PBST for 1 hr, equilibrated in 50% glycerol/50% Ringer’s solution (5 mM HEPES, pH 7.3; 130 mM NaCl; 5 mM KCl; 2 mM MgCl₂; 2 mM CaCl₂) for 15 min or overnight at 4°, then mounted in 2% nPropyl-gallate in 90% glycerol. Primary antibodies were diluted in normal donkey serum as follows: mouse antibody against Antp (1:50; DSHB, 8C11), rabbit antibody against Odd skipped (1:400; gift from James Skeath, Washington University School of Medicine) chick antibody against GFP (1:1500; Aves Labs, GFP-1020), rabbit antibody against RFP (1:1000; Abcam, ab62341), mouse antibody against Fasciclin 3 (1:50; DSHB, 7G10), rabbit antibody against Mef2 (1:1000; DSHB), guinea pig antibody against Odd skipped (1:1200; gift from John Reinitz, University of Chicago), rabbit antibody against Bin (1:100; gift from Eileen Furlong, EMBL), mouse antibody against Slit (1:200; gift from Greg Bashaw, University of Pennsylvania), mouse antibody against Robo1 (1:200; gift from Greg Bashaw, University of Pennsylvania), chick antibody against LacZ (1:1000; Abcam, ab9361-250). Secondary antibodies (Alexa Fluor 488, Cy3, Cy5, and Alexa Fluor 647; Molecular Probes or Jackson ImmunoResearch) were all used at 3.75 µg/mL, for 1 hr.

After dechorionation, embryos were selected by stage and transgene expression, using a stereoflouorescent microscope, transferred to a piece of agar, and then oriented in a line with the ventral/ventrolateral surface facing up – hanging off the edge of the agar. A heptane-glue mixture was dried as a strip on a glass slide, and that surface touched to the embryos for transfer (now dorsal/dorsolateral surface of embryos faces up). 3 µL of halocarbon oil was added atop the embryos. Bridging coverslips were glued to the slide on either side of the line of embryos, then the main coverslip was laid atop the bridging coverslips. We imaged every 5–10 min for 2–4.5 hr; Z-stacks spanned 25–40 µM with 0.3–1.0 µM step sizes.

Fixed embryos were imaged on a Zeiss Axio Imager with ApoTome using a 40 x, 1.2 NA water immersion objective or a 20 x, 0.8 NA objective; z-steps were 0.5–1.0 uM. All live-imaging except for alary muscle imaging occurred with a CrestOptics X-Light V3 spinning disk confocal microscope using a 60 x, 1.3NA silicone immersion objective; images were captured with two pco.edge 4.2 bi sCMOS cameras operated by VisiView (Visitron) software. Alary muscles were live-imaged with an IX7 Olympus spinning disk confocal using a 63 x, NA 1.2 water immersion objective and captured with an EMCCD camera (Hamamatsu photonics, model C9100-13) controlled by MetaMorph software.

We used immunostains for the accepted markers, Antp and Odd, to identify PSC cells. Sometimes images contained Antp + and Odd + cells in epidermal stripes anterior to the LG; these were not considered PSC cells. An embryo was scored as having ‘normal’ PSC positioning if both PSCs were (1) coalesced within one nuclear diameter of one another, (2) adjacent to the dorsal vessel, and (3) at the same dorsal-ventral position as the posterior-most cells of the lymph gland. Abnormal PSC positioning presented as a range of phenotypes including PSCs dispersed into multiple groups, and PSCs that were coalesced but not located at the LG posterior. Fisher’s Exact test was used for statistical analysis and p<0.05 was considered statistically significant.

We considered a cell to be a PSC cell if it was co-labeled by Antp and Odd and was not located in epidermis. Number of PSC cells was recorded separately for the left and right PSCs of each embryo, and the total number of PSC cells per embryo was plotted. The Mann-Whitney test was used to determine a non-significant difference with p>0.05.

For 9 sibling controls and 10 Vm-ablated embryos, a normalized Bin fluorescent intensity was calculated for three different regions of st11 Vm founder cells with Bin-stained nuclei and Fas3-stained membranes. For st11 embryos imaged laterally, Fas3 labels the 2 lateral surfaces of a given Vm founder cell membrane. ImageJ software was used to extract fluorescent intensity values for regions of interest including large non-positive background regions in the embryo, founder cell nuclei, and founder cell membranes. Bin intensity for a given region was calculated by normalizing the background subtracted Bin level of a single representative nucleus to the average background subtracted Fas3 level of at least the two lateral surfaces of that same cell; most often an average Fas3 level for a given region was calculated based on the membranes of 2–4 founder cells. The normalized Bin fluorescent intensity for a given region is plotted. The Mann-Whitney test was used for statistical analysis to determine a significant difference of p<0.05.

Two independent UAS-controlled RNAi’s targeting different regions of Exon 4 of Slit along with UAS-controlled Dcr2 were driven by tinCΔ4-GAL4. Knockdown was evident at st14, the same stage when Slit expression becomes discernable in control CBs.