Introduction

The diversity of cell types in the nervous system arises from a small pool of neural stem cells (NSCs). Decades of work in Drosophila show that NSCs use a hierarchical regulatory logic: positional cues endow NSCs with distinct molecular identities and lineage potentials, while a temporal gene cascade determines the production of different neuron types over time. NSCs activate neuron-specific terminal selector (TS) genes by integrating these spatial and temporal inputs along with Notch signalling. TSs then drive effector gene programs that implement neuronal morphology, connectivity, and neurotransmitter identity (1–3).

This has been best elucidated in the embryonic ventral nerve cord (VNC) and the optic lobe. In the VNC, 30 neuroblasts (NBs) per hemisegment each generate a lineage with a unique identity that is determined at the time of delamination (4–8). This identity reflects the NB’s neuroectodermal origin, where intersecting anterior–posterior (A–P) and dorsoventral (D–V) patterning systems create distinct domains of TF expression (2, 3, 9) (Fig. 1A,B). NBs inherit these spatial TFs, which specify lineage identity. For example, the A–P gene gsb is expressed in row 5 of the neuroectoderm and required for row-5 identity: its loss transforms NBs to a row 3/4 fate, while misexpression induces the converse transformation (10). Other STFs act non-autonomously: wingless (wg) shapes the identity of adjacent rows without directly specifying row 5 (11). More posteriorly, en defines rows 6 and 7: en mutants lack NB6-4 and NB7-3, while en misexpression duplicates NB7-3 identity (12–14). D–V patterning factors, including vnd, ind, and msh, similarly define medial, intermediate, and lateral NB columns. Gain- and loss-of-function studies demonstrate that these D–V genes also respecify NB columnar identity (15–18). Together, these intersecting spatial cues generate 30 distinct NB identities per hemisegment, each with a unique developmental potential.

Fig. 1.

In the optic lobe, an analogous spatial logic partitions the neuroepithelium of the outer and inner proliferation centers (OPC/IPC) into A–P domains by Vsx, Optix, and Rx (split by Dpp and Wg and into D–V compartments by Spalt/Salm and Disco/Hh) (1, 19–23). These factors establish progenitor territories that seed distinct neuronal outputs across the medulla, lobula, and lobula plate (20–23). For example, Vsx-domain progenitors generate Pm3 neurons, dorsal and ventral Rx territories produce Pm1 and Pm2 respectively, whereas Optix-domain equivalents are eliminated by apoptosis (20). As in the VNC, manipulating spatial inputs reassigns territories and shifts lineage output. For example, expanding the Rx domain biases neural fates towards Pm1/2 output, and mutating disco, which specifies ventral fates, results in loss of ventral medulla neuron identities (20, 22).

NBs are also patterned in time. Each NB sequentially expresses a cascade of TTFs, creating successive competence windows during which different neurons are born (Fig. 1C) (1, 2, 19, 24, 25). In the embryonic VNC, the canonical cascade proceeds through Hunchback (Hb) → Kruppel (Kr) → Pdm → Castor (Cas) → Grainyhead (Grh) (26–28). In the optic lobe, this cascade consists of Hth → Opa → Erm → Ey → Hbn → Scro → Slp → D → BarH1 → Tll (29–34). While the identity of the temporal factors varies across systems — including the central brain, type II lineages, and in the postembryonic VNC and brain — the strategy is conserved: sequential expression of TTFs subdivides neurogenesis into discrete temporal windows, each producing neuron types appropriate to that time point.

In keeping with this, manipulating temporal factors respecifies the temporal identity of the neurons. In the optic lobe, loss of function of Ey, Slp, or D stalls the temporal sequence and the neurons from those windows are lost; misexpressing any one of them biases production toward that window’s fates (31). For example, in the early Hth/Bsh window, bsh or hth loss eliminates Mi1 neurons, and Bsh overexpression induces Mi1-like neurons (31, 35, 36). Similarly, in the embryonic VNC, Hb, the first in the temporal cascade, is both necessary and sufficient for the earliest-born neurons in multiple NB lineages. Its misexpression prolongs early competence windows and induces supernumerary early-born neurons, while its loss abolishes these fates (26, 27, 37–40).

Importantly, the identity of a neuron from a given time window differs across lineages. For example, Hb+ neurons are: U1 in NB7-1, RP2 in NB4-2, EW1 in NB7-3, and CCAP in NB3-5 — implying that temporal information alone is insufficient to specify neuronal identity (24). Instead, Hb indeed any TTF — must act in a spatially defined context to specify neuronal fate.

The mechanism by which spatial and temporal cues are integrated to generate such context-dependent outcomes remains unclear. Two mechanistic models have been proposed. In direct regulation, STFs and TTFs (Hb) act concurrently: both are present at the same time in the same NB, and co-occupy TS enhancers to drive activation. In epigenetic regulation, the action is sequential: an STF first engages the enhancers to establish an accessible state, and only later does the TTF bind and activate transcription (Fig. 1D).

Regardless of the model, we refer to the cis-regulatory loci that integrate the STF and TTF inputs as “sites of integration” (SoIs). Operationally, SoIs are enhancer elements that: (i) are specific to a particular NB lineage, and (ii) receive input from both the lineage-defining STF(s) and the TTF. Under direct regulation, SoIs are already in a receptive state and show concurrent STF–TTF occupancy. Under epigenetic regulation, SoIs bear evidence of prior STF action that establishes lineage-specific competence for later TTF recruitment (Fig. 1D).

Here, we test these models using the two well-characterized embryonic lineages — NB5-6 and NB7-4 — and their putative STFs, Gsb and En respectively. Engrailed (En) persists in NB7-4 whereas Gooseberry (Gsb) is transient and precedes Hunchback (Hb) in NB5-6, suggesting that NB56 must use epigenetic regulation, while either could operate in NB7-4. We therefore used chromatin engagement as the discriminator between direct and epigenetic integration. If integration is epigenetic, the lineage STF must engage lessaccessible chromatin to establish NB-specific sites of integration (SoIs) — enhancers rendered competent for later TTF recruitment. If regulation is direct, the STF need not engage less-accessible chromatin; concurrent STF–TTF binding on pre-accessible enhancers suffices. Guided by these predictions, we asked whether En in NB7-4 shows evidence of priming and whether Gsb in NB5-6 engages and remodels less-accessible loci—and whether such priming licenses later Hb recruitment.

Prior work mapping Hb and chromatin state in identified NBs showed that Hb binds distinct, lineage-specific loci that coincide with NB-specific open chromatin (SoIs) — and that these differences track the distribution of the STF Gsb — supporting a priming step that licenses later Hb binding (41). However, in that study, Gsb occupancy was obtained from available whole embryo ChIP-seq data and so NB-specific occupancy was lacking. Here we profile STF occupancy and chromatin accessibility in each lineage in vivo and test the sufficiency of STFs to reprogram accessibility and guide Hb binding. In NB7-4, En binds only pre-accessible loci, ruling En out as the priming factor and aligning with direct regulation on a pre-opened template. In NB5-6, Gsb binds both open and less-accessible chromatin and, when misexpressed in NB7-4, remodels chromatin genome-wide: it directly reduces accessibility at its binding targets, including at endogenous NB7-4 SoIs, where Hb occupancy is correspondingly weakened. It also indirectly renders some sites more accessible and such opened sites rarely recruit Hb. Across both lineages, the most accessible SoIs are co-bound by the STF and Hb, consistent with cooperative enhancer activation once chromatin has been appropriately primed.

Together, our findings support a unified model in which spatio–temporal integration occurs in two steps: NB-specific STF combinations first prime lineage-specific chromatin and then, together with lineage-specific cofactors, recruit and restrict Hb to generate NB-specific enhancer landscapes. This, in turn, supports a model where lineage identity is determined by unique combinations of STFs—the ‘STF code’—whose members partition the molecular functions of chromatin priming and Hb patterning that confer spatial identity within a lineage.

Results

Generation of functional Dam:Gsb and Dam:En to determine NB-specific STF occupancy

To determine STF occupancy in vivo in an NB-specific manner, we used Targeted DamID (TaDa), which has been employed previously to measure the Hb TTF’s occupancy in neuroblasts (NBs) (41). We generated UAS-LT3-Dam:Gsb and UAS-LT3-Dam:En fly lines, in which an upstream mCherry ORF ensures low expression of Dam:Gsb or Dam:En from a downstream ORF (see Methods, Fig. S1A). This configuration minimizes toxicity associated with high Dam levels (42).

We first confirmed that Dam:Gsb and Dam:En were not toxic to cells. Their ubiquitous expression under Da-Gal4 caused no embryonic lethality and no denticle belt defects. However, this lack of toxicity might reflect non-functional TFs when fused to Dam. We therefore tested the functionality of these fusion proteins. We performed whole-embryo Dam:Gsb profiling (Da-Gal4>UAS-LT3-Dam:Gsb) and compared it to published Gsb ChIP data (43). Three biological replicates were generated, alongside UAS-LT3-Dam controls (DamOnly), and processed as described (see Methods). Dam:Gsb and DamOnly replicates correlated strongly within their own sets but poorly with each other (Fig. S1B), confirming reproducibility. Dam:Gsb also reproducibly marked known Gsb targets (wg, gsb-n, prd) (Fig. S1C), and signals were enriched at Gsb-ChIP peaks but not at ChIP peaks of unrelated TFs (Fig. S1D). Dam:Gsb peaks should be enriched for Gsb binding motifs. Although a canonical Gsb motif is not well-defined, Gsb contains both homeodomain and paired DNA-binding domains (44). A de novo motif search revealed enrichment of a paired-homeobox composite motif in Dam:Gsb peaks relative to background (Fig. S1E). Similarly, Dam:En data were reproducible, showed occupancy at established En targets, and were enriched for the canonical En motif (Fig. S1F-H).

Together, these experiments confirm that Dam:Gsb and Dam:En retain DNA-binding activity, are non-toxic to cells, and accurately report genomic occupancy, providing validated tools to interrogate STF binding in NB5-6 and NB7-4.

Assaying spatial TF occupancy in identified neuroblasts

STFs such as Gsb and En are expressed broadly in the embryonic neuroectoderm, yet their lineage-specific activities are executed within individual NBs. This creates a major challenge for profiling their genomic occupancy: each NB comprises only a small number of cells, and STF expression is often transient. Previous studies relied on whole-embryo ChIP data (45), which cannot resolve NB-specific binding. To overcome this limitation, we combined TaDa with NB-specific Gal4 drivers to assay STF binding in two identified lineages: Dam:Gsb in NB5-6 and Dam:En in NB7-4 (Fig. 2A-B). We collected data during stages 10–12, corresponding to the Hb competence period when SoIs are active. Despite sampling very few cells over this short time window, replicate Dam:Gsb and Dam:En profiles were highly reproducible and detected occupancy at established Gsb and En target loci (Fig. 2C-F). Dam-only controls confirmed previous findings that NB5-6 and NB7-4 possess distinct chromatin landscapes, with many loci uniquely accessible in one NB but not the other (1933 loci at FDR ≤ 0.05; Fig. 2G-H). Together, these results validate that TaDa can be used to reliably profile STF occupancy in individual NBs, providing the necessary tools to test how spatio-temporal integration occurs within neuroblasts.

Fig. 2.

En binds only accessible chromatin in NB7-4

The two models - direct or epigenetic regulation - make clear predictions for STF occupancy in the context of the chromatin within NBs. While STF occupancy at less accessible chromatin is necessary under the epigenetic regulation, it is dispensable under direct regulation.

To test whether NB7-4 follows direct or epigenetic regulation, we asked whether En binds only at pre-open sites (consistent with conventional TF behaviour) or also engages less accessible chromatin to create SoIs (as predicted by epigenetic priming).

We first focussed our analyses on the previously generated SoIs (41). These sites were identified by performing a differential analysis on Hb binding in NB5-6 and NB7-4, and were confirmed to be differentially accessible in the two NBs as previously reported (see Methods, Fig. S2, Fig. 3A). As expected, En was enriched at NB7-4-specific SoIs, where Hb uniquely binds in this lineage (Fig. 3B). However, there was no detectable En signal in NB7-4 at sites known to be NB5-6-specific SoIs (which remain inaccessible and Hb-free in NB7-4) (Fig. 3C). Thus, En associates with SoIs only once they are already open and competent for Hb binding, but does not engage closed SoIs that might require priming.

Fig. 3.

This argues against En’s role as an STF that primes lineage-specific chromatin by binding both accessible and less accessible regions to establish SoIs. Before concluding that En does not act via the epigenetic model, we examined its binding genome-wide, relative to chromatin accessibility (Fig. 3D). Because we had already validated the Dam:En tool and defined genome-wide NB-specific accessibility landscapes in NB7-4 and NB5-6 using Dam-only controls, we could confidently classify NB7-4 chromatin into loci that were uniquely accessible or less accessible through differential analysis of Dam binding between the two NBs: sites uniquely Dam-bound in NB7-4 were accessible and those that were uniquely open in NB5-6 were less accessible in NB7-4 (see Methods, Fig. S3, Fig. 3D).

At the genome-wide level, En was strongly enriched at accessible chromatin (Fig. 3E) and showed no detectable binding at less accessible sites (Fig. 3F). Thus, consistent with our earlier findings, En’s binding is confined to pre-open regions. To test this relationship more directly, we compared chromatin accessibility at En-bound versus En-unbound loci genome-wide. A differential analysis between En (NB7-4) and Gsb (NB5-6) yielded ‘En-bound’ sites in NB7-4, that were not Gsb-bound in NB5-6. Conversely, ‘En-unbound’ sites corresponded to those that were Gsb-bound in NB5-6, but not En-bound in NB7-4 (see Methods, Fig. S4, Fig. 3G). En-bound regions displayed markedly higher accessibility than unbound ones (Fig. 3H-I), demonstrating that accessibility is not only correlated with but predictive of where En will bind. In other words, En binding is conditional on chromatin being pre-open. This pattern is consistent with conventional TF activity and argues against a pioneer role for En.

Taken together, these results show that En binds exclusively to accessible chromatin and associates with SoIs only after they are open, indicating that En does not act as a pioneer factor. Instead, it operates on enhancers that have already been primed - likely by other lineage-specific pioneer factors. This suggests that NB7-4 employs an ‘STF code’ in which different members have distinct functions — unknown pioneer factors establish its SoIs, while non-pioneer factors like En read the pre-open state and likely collaborate with Hb to activate lineage-specific enhancers (tested below).

Gsb binds both open and closed chromatin in NB5-6

Having established that En in NB7-4 binds pre-accessible chromatin, we next evaluated whether NB5-6 follows direct or epigenetic regulation by asking if Gsb, like En, binds pre-accessible chromatin or binds broadly across accessible and less accessible regions to establish SoIs (epigenetic priming). At NB5-6-specific SoIs (Fig. 4A), Gsb occupancy was enriched as expected (Fig. 4B). However, Gsb signal in NB5-6 was also enriched at sites identified to be NB7-4-specific SoIs, which remain less accessible and Hb-free in NB5-6 (Fig. 4C). Gsb occupancy at NB7-4-specific SoIs indicates that Gsb binding alone does not determine where Hb will be recruited.

Fig. 4.

This suggested that Gsb has the ability to bind less accessible chromatin. To confirm this, we extended our analysis to the whole genome and examined Gsb’s occupancy in the context of chromatin accessibility. We used the previously generated Dam:Gsb binding data in NB5-6 and the already defined categories of chromatin loci from the previous sections - accessible in NB5-6 and less accessible in NB5-6. As in the previous section, differential analysis of Dam binding between the two NBs identified Dam-bound sites enriched specifically in NB5-6 as accessible, and NB7-4-enriched sites as less accessible in NB5-6 (see Methods, Fig. S3, Fig. 4D). At the genome-wide level, Gsb was enriched at accessible chromatin as expected (Fig. 4E). Importantly, we also detected some Gsb occupancy at less accessible regions (Fig. 4F). This property distinguishes Gsb from canonical TFs and is consistent with pioneer-like activity.

To confirm this, we compared chromatin accessibility at Gsb-bound versus unbound loci, which were determined as described earlier via differential analysis between Gsb (NB5-6) and En (NB7-4): enriched occupancy in NB5-6 yielded ‘Gsb-bound’ sites and ‘Gsb-unbound’ sites corresponded to those that were En-bound in NB7-4, but not Gsb-bound in NB5-6 (see Methods, Fig. S4, Fig. 4G). Genome-wide, accessibility was significantly lower at Gsb-bound regions relative to unbound ones (Fig. 4H-I), consistent with Gsb engaging both accessible and less accessible chromatin.

Together, these results meet a key prediction of the epigenetic model: Gsb binds both open and closed chromatin and occupies SoIs. However, because many Gsb-bound sites, including SoIs (those defined as NB7-4-specific), remain inaccessible and Hb-free in NB5-6, Gsb binding is not sufficient to recruit Hb. We infer that Hb recruitment in NB5-6 requires Gsb together with additional NB-specific cofactors rather than arising from Gsb–Hb direct co-occupancy alone. These findings also point to NB5-6 using an STF code with role partitioning: Gsb primes chromatin to establish SoIs, while other lineage-restricted factors recruit Hb and activate lineage-specific enhancers.

Co-binding with Hb marks highly accessible sites

Our analyses of Gsb in NB5-6 and En in NB7-4 suggest a division of labour within each lineage’s STF code: some factors, like Gsb, may act as pioneers to prime chromatin, while others, like En, collaborate with Hb at enhancers that are already open. If enhancer activation requires both priming and TTF engagement, then loci co-bound by Hb and an STF-code component should represent the most active regulatory sites. To test this, we analyzed three genome-wide classes of loci — STF-only, Hb-only, and STF+Hb co-bound (Fig. 5A-B) — and compared their chromatin accessibility. For each TF (Gsb in NB5-6; En in NB7-4; Hb in both), the top 5000 highest-confidence peaks covering known genomic targets were used. To define STF-only sites, we subtracted Hb-bound peaks that overlapped with STFs; the same approach was used to obtain Hb-only sites in each NB. Co-bound sites were defined as those with an overlap of at least 90%. In both NB5-6 and NB7-4, accessibility was lowest at Hb-only sites, intermediate at STF-only sites, and highest at STF+Hb cobound loci (Fig. 5C-H). This analysis demonstrates that dual occupancy with Hb correlates with maximal accessibility at SoIs. This suggests that effective enhancer activation in these lineages proceeds in two steps: chromatin priming by STF-code components, then assembly of Hb with an STF-code partner. Co-binding marks the activated enhancers.

Fig. 5.

Gsb remodels chromatin when ectopically expressed

Having established that Gsb in NB5-6 can bind both open and closed chromatin, consistent with pioneer-like potential, we next asked whether Gsb is sufficient to actively remodel chromatin accessibility. Binding to closed loci is a hallmark of pioneers, but our proposed two-step model makes a stronger prediction: that an STF should not only recognize compact chromatin but also remodel it to generate accessible sites competent for later TTF binding and activation. Addressing this prediction requires moving beyond correlative occupancy analyses to a functional test of whether Gsb can alter chromatin state when introduced into a lineage where it is not normally expressed.

To this end, we ectopically expressed Gsb ubiquitously using a Hs:Gsb construct during the Hb-competence window and assayed chromatin accessibility in NB7-4, a lineage where Gsb is absent (see Methods, Fig. S5, Fig. 6A). This provided a context to assess whether Gsb could pioneer new sites of accessibility. Ectopic Gsb binding produced considerable remodelling: about 150 previously closed loci gained accessibility, while a similar number of normally open sites became less accessible (Fig. 6B,C). Crucially, only the decreasing sites overlap independent Gsb-binding datasets (whole-embryo Dam:Gsb), whereas the increasing sites do not (Fig. 6D,E). This pattern suggests that Gsb’s direct effect is to reduce accessibility at its binding targets, while increases likely reflect indirect consequences of the perturbation.

Fig. 6.

We then asked whether such remodelling was sufficient to fulfil the second step of the model: recruitment of Hb to the newly accessible loci. To test this, we had simultaneously profiled Hb occupancy in the same embryos (see Methods). Despite Gsb-induced changes in chromatin accessibility, only a small fraction of these newly opened sites gained Hb binding, and even then the enrichment was weak (Fig. 6F,G).

We next focussed our analysis on established SoIs in NB7-4 and asked how ectopic Gsb affects them. At loci corresponding to NB5-6 SoIs, which are normally less accessible and Hb-free in NB7-4, Gsb misexpression neither increased accessibility nor recruited Hb: these sites remained closed and Hb-depleted (Fig. 6H, J). Thus, Gsb on its own is unable to install NB5-6-like SoIs in NB7-4. In contrast, at endogenous NB7-4 SoIs, where chromatin is normally highly accessible and strongly bound by Hb, ectopic Gsb led to a marked reduction in accessibility (Fig. 6I). We also observed a reduction in Hb occupancy at these sites (Fig. 6K). Ectopic Gsb therefore partially antagonizes the native NB7-4 SoIs rather than creating new ones, consistent with its genome-wide role.

Together, these experiments demonstrate that Gsb is sufficient to remodel chromatin when misexpressed, providing direct functional evidence that it has pioneer-like activity. Importantly, our data suggest that Gsb acting on its own tends to close rather than open chromatin, including at SoIs. Furthermore, Gsb-induced accessibility changes only partially alter Hb binding, showing that enhancer activation requires more than chromatin opening. Taken together, these findings support and refine our earlier conclusion that productive opening and Hb recruitment arise from the NB-specific STF code, in which Gsb operates with additional lineage-specific factors to establish bona fide SoIs.

Manipulating Gsb alters NB lineage identity

Having shown that Gsb can function as a pioneer to remodel chromatin, we next asked whether this activity translates into changes in NB lineage identity. The two-step model predicts that pioneering activity by STFs establishes the chromatin landscape on which temporal factors act, thereby constraining lineage potential. If this is correct, then manipulating Gsb levels should directly impact NB fate.

We first depleted Gsb in the early embryo using Vasa-Cas9, which is known to be expressed in somatic cells of the early embryo (51), together with a gsb-targeted guideRNA. Embryos lacking gsb were embryonic lethal and exhibited axonal phenotypes consistent with previously described gsb loss (Fig. S6A-B). Sequencing such embryos at the time of NB delamination verified a mosaic but high rate of loss of Gsb function, confirming that we could induce gsb loss-of-function in early embryos (Fig. S6C-F).

In wild-type embryos, we could identify NB5-6-born Tv neurons, which express both Ap and Eya markers (52) (Fig. 7A, C-F). In gsb loss-of-function embryos, these NB5-6 markers along with the NB5-6-specific Lbe-K enhancer were lost, demonstrating that Gsb is necessary for maintaining NB5-6 identity (Fig. 7B, G-I). Ap-positive neurons are also born from MP2 (in row 3) and NB4-3, named dMP2 and dAp neurons, respectively (53) (Fig. 7A, C, J-L). We noted twice the usual number of dMP2 neurons in several hemisegments, while dAp neuron numbers remained unaffected (Fig. 7 M-O). In addition, there was a concomitant ectopic expression of the NB3-5-specific marker, Ems (Fig. 7P-Q). Since NB5-6-specific Eya- and Ap-positive cells arise only in the thoracic hemisegments, we limited our quantification of these markers to these hemisegments alone. Quantification of these markers across hemisegments confirmed this reciprocal fate switch - suppression of NB5-6 identity and gain of NB3-5 fate (Fig. 7R-T), demonstrating that Gsb is indispensable for maintaining NB5-6 identity and preventing re-specification toward an NB3-5 fate. These findings also confirm earlier reports that Gsb specifies row 5 and suppresses row 3 fates (10).

Fig. 7.

To test whether Gsb is sufficient to impose NB5-6 identity, we broadly misexpressed Gsb at the time of NB delamination and assayed the same set of molecular markers (Fig. 8A). Ectopic Gsb expression resulted in a loss of NB3-5 identity-a proportion of hemilineages showed a complete loss in NB3-5-reporter R59E09>GFP and Ems (Fig. 8B-C). Notably, Ems also marks NB3-3, another row 3 NB (6). Therefore, a complete loss of Ems from an entire hemisegment suggests a broad disruption of row 3 identity following ectopic Gsb expression, consistent with previous findings (10). Broad Gsb misexpression also induced a gain of NB5-6 identity, marked by Lbe>GFP. In several thoracic hemisegments, we observed supernumerary Eya- and Ap-positive neurons, consistent with a gain in NB5-6T-born Tv neurons (Fig. 8D-E). However, in only a small proportion of the hemisegments was there a concomitant gain in NB5-6-identity and loss of NB3-5 identity. Quantification confirmed these findings (Fig. 8F-J), indicating that while Gsb is sufficient to repress alternative fates, it is not always enough on its own to fully impose NB5-6 identity.

Fig. 8.

Taken together, these experiments establish Gsb as the critical pioneer factor in establishing NB5-6 identity and suppressing NB3-5 fate. However, in both cases the transformations were incomplete: Gsb overexpression did not consistently install NB5-6 identity, and Gsb loss did not fully erase it. This parallel outcome suggests that while Gsb initiates chromatin remodelling and biases fate, complete lineage specification requires additional components of an NB5-6 STF code—factors that both stabilize identity and collaborate with temporal regulators like Hb.

Discussion

Neural stem cells must integrate spatial and temporal information to produce lineage-appropriate time-specific neurons. Two models have been proposed to explain how this could occur: direct regulation model, in which STFs and TTFs cooccupy and activate enhancers of TS genes to deploy effector programs, and an epigenetic regulation model, in which early STFs establish open chromatin that later permits TTF binding and enhancer activation. Our results in two identified neuroblasts support a mechanism that contains features of both models. We propose that each NB carries a unique STF code — a combination of spatial TFs that confers lineage identity — whereby some members prime lineage-specific chromatin, while others collaborate with the TTF to recruit and activate lineage-specific enhancers.

The functional manipulations of gsb (both loss- and gain-of-function) are consistent with this division of labour: priming shifts chromatin accessibility and biases lineage identity, but full enhancer activation and therefore complete lineage transformation would require Hb recruitment by the appropriate NB-specific code. It is, however, possible that other neuroblasts or time windows use other strategies for integrating position and time — including TTF-led chromatin remodelling - the TTF Grainy head is a known pioneer (54) - or direct co-occupancy on DNA.

Gsb as a context-dependent pioneer

Our data firmly establish Gooseberry (Gsb) as a pioneer-like regulator in NB5-6. Genome-wide TaDa profiling shows that Gsb binds both accessible loci and regions that remain relatively inaccessible in this lineage. Ectopic expression experiments further demonstrate that Gsb is sufficient to remodel chromatin in the non-cognate NB7-4 in both directions: closing previously open sites and generating newly open ones. Together, these observations are consistent with context-dependent pioneer activity (55).

Direct pioneer activity has not been shown previously for Drosophila Gsb in neuroblasts, but Gsb is a Pax3/7-class TF that specifies row-5/6 NB identity in the neuroectoderm, placing it at the top of the spatial cascade that patterns these lineages (10). Within the Pax family, Gsb (and Prd) are Pax3/7 homologues by sequence and function (56, 57). In vertebrates, Pax7 has well-documented pioneer activity: it can engage nucleosomal DNA and deploy de novo enhancer repertoires, but efficient chromatin opening often requires a partner (for example Tpit, in the pituitary) (58–60). Pioneering and chromatin-opening functions of Pax7/Pax3 have also been documented in the myogenic lineages (61–63). In an oncogenic context, PAX3–FOXO1 fusion TF reprogrammes chromatin to install myogenic super-enhancers (64).

Pioneer binding does not guarantee activation. Across systems, pioneers can either seed opening (via co-activators and remodellers) or contribute to repression (via corepressors/HDACs), with the outcome set by partner availability and local chromatin context (55). For Pax7 specifically, binding and opening are separable: Pax7 can recognize heterochromatin on its own, but robust accessibility requires Tpit, illustrating why a pioneer may open some targets yet appear to close or dampen others when the right partners are missing (59). This division of labour aligns with our finding that Gsb remodels chromatin bidirectionally upon misexpression. More broadly, vertebrate Pax3/7 studies emphasize co-factor dependence, mirroring NB5-6, where Gsb-generated accessibility is insufficient for Hb recruitment.

En as a cofactor-dependent effector at pre-accessible enhancers

En in NB7-4 provides a counterpoint to Gsb. It binds almost exclusively to pre-accessible loci, with no detectable association with less accessible chromatin. This is in keeping with the wider literature: to our knowledge, neither Drosophila En nor its vertebrate homologues (EN1/EN2) have been shown to bind nucleosome-occluded DNA and initiate chromatin opening — the operational definition of pioneer activity (55). Thus, En lacks pioneer activity and likely acts as a recruiter or stabilizer with Hb at lineage-specific SoIs. En+Hb co-bound sites show the highest accessibility, consistent with cooperative enhancer activation. Because En does not open chromatin de novo, a different factor must prime NB7-4 enhancers - identifying this will be key to understanding lineage specification of NB7-4.

En can both activate and repress its target genes often via chromatin-modifying co-factors. As a repressor, En’s eh1 motif recruits Groucho/TLE co-repressors, which in turn recruit HDAC/Rpd3, providing a route to local deacetylation and compaction (65–68). As an activator, En can partner with CBP/p300 (Nejire) at a regulatory element in vivo, linking En-dependent activation to histone acetylation (69). En can also act as a bona fide activator at specific targets (e.g., polyhomeotic) (69–71).

In summary, the best-supported chromatin mechanism for En is cofactor-mediated modulation of already open enhancers — activating via CBP/p300 and repressing via Gro/TLE — rather than pioneering access to closed chromatin, which fits our NB7-4 data.

The STF code for each NB

We invoke an “STF code” because of two lines of evidence: classic work shows that NB identity is combinatorial, set by orthogonal AP and DV patterning systems, and our data that reveals a division of labour among STFs (priming vs. effector roles), implying that no single STF suffices. Foundational mapping studies by Urbach and Technau explicitly framed NB identity as unique combinations of TFs (72). They linked 40–46 markers to individual brain NBs, effectively cataloguing NB-specific codes without yet resolving their mechanistic roles.

This suggests a practical definition: an NB’s STF code is the minimal combination of STFs sufficient to confer lineage identity by delimiting the enhancer repertoire and specifying how the temporal cascade engages it—whether via STF-driven priming and TTF recruitment, direct co-occupancy on pre-accessible DNA, or, in some contexts, TTF-led chromatin remodelling with STFs selecting targets.

A corollary is testable: closely related NBs should be inter-converted by swapping the relevant code components. An extreme example of this was shown in the central brain where manipulation of a single TF, otd, in a pair of closely related NB lineages could completely change their identity rewiring both morphology and function (73). In general, however, the STF code predicts partial switches with manipulations of single factors (as seen for gsb or vnd) and more complete identity changes when multiple code members are comanipulated. Systematic code mapping across NBs — extending the Urbach and Technau atlases with perturbation and chromatin readouts — should reveal how modest changes in code composition generate the observed diversity of neuronal lineages.

Relation to vertebrate neural tube

Shared principles are seen in vertebrates. For example, spatial patterning has been exquisitely mapped in the spinal cord. Opposing morphogen gradients (ventral Shh, dorsal BMP/WNT) establish discrete progenitor domains along the dorsoventral axis, while Hox genes specify columnar and pool identities along A–P levels to diversify motor and interneuron classes (74). Temporal patterning, initially well defined in the developing cortex, has more recently has been generalized across vertebrate CNS regions, including the spinal cord. A shared temporal transcription factor code has been described that unfolds across progenitors to coordinate neuronal subtype production in time, complementing spatial patterning inputs (75).

These vertebrate studies raise a natural question: how are spatial and temporal cues integrated? A recent study suggests a distinct strategy for integrating spatial and temporal information. Zhang et al. report a global temporal chromatin program unfolding across progenitor domains that operates in parallel to spatial patterning: temporal regulators orchestrate the opening of regulatory elements over time, and spatial determinants act on the set of elements that become available. They describe this as a “chronotopic” integration strategy and it predicts SoI-like elements in vertebrates where temporal state broadly reconfigures accessibility and competence, onto which spatial TFs then act (76).

Materials and Methods

All fly stocks are listed in Table 1. All antibodies used are listed in Table 2.

Table 1.

Table 2.

Fly Stock Maintenance

Fly stocks were maintained on a standard cornmeal medium. The recipe consisted of the following per litre: 80g corn flour, 20g glucose, 40g sugar, 15g yeast extract, 4mL propionic acid, 5mL p-hydroxybenzoic acid methyl ester (in ethanol), and 5mL ortho-butyric acid. Flies were raised at 25°C under a 12h:12h light-dark cycle. The virginator and hsGsb stocks were grown at room temperature.

Generation of Dam:Gsb, Dam:En, and gsb-guideRNA fly lines

To generate the gsb-guideRNA line, Drosophila gsb-guideRNA sequences were cloned into the tandem guideRNA expression vector pCFD4 ((77)). The gRNA target sequences were amplified via PCR using primers, and the resulting PCR product was assembled into the BbsI-digested pCFD4 backbone using Gibson assembly.

TaDa and CATaDa Experiments

Verification of Dam:Gsb and Dam:En

To assess whole-embryo occupancy of Gsb and En, Dam fusion constructs were expressed using the ubiquitous Da-Gal4 driver. 1000-1600 females of w-;;Da-Gal4 were crossed to a similar number of males of w-;;UAS-LT3-Dam:Gsb for Gsb occupancy, w;;UAS-LT3-Dam:En for En occupancy, and w-;;UAS-LT3-DamOnly for chromatin accessibility.

After 16–24 hours of mating, adults were transferred to embryo collection cages. Collection plates were prepared using 2% agar and 2% sucrose in 60–100 mm Petri dishes, and yeast paste was applied to stimulate egg laying. Embryos were collected every 4 hours (0–4 h AEL) and then aged for 12 hours (12–16 h AEL; Stages 15–16) at 25°C.

Following ageing, embryos were rinsed with distilled water and dechorionated in bleach for 2 minutes, or until complete removal of the chorion was visually confirmed. Embryos were then washed thoroughly with distilled water, dried, weighed, and stored in 1.5mL tubes at –20°C. For each biological replicate, a minimum of 30 mg of embryos was collected.

Determination of NB-specific STF occupancy and chromatin accessibility

To determine Gsb occupancy and chromatin accessibility in NB5-6, we crossed 4000 males of w-;Lbe-k-Gal4 to 4000 females of w-;;UAS-LT3-Dam:Gsb or w-;;UAS-LT3-DamOnly. Similarly, to determine En binding and chromatin accessibility in NB7-4, 4000 males of w-; 19B03[AD];18F07[DBD] were crossed to 4000 females of w-;;UAS-LT3-Dam:En or w-;;UAS-LT3-DamOnly. Embryos were collected for 2h (0-2h AEL) and then aged for 6h (6h-8h AEL, Stage 11-12) at 25°C. We collected 4 replicates of at least 30mg of embryos each for NB5-6-specific Dam:Gsb and DamOnly, and 3 replicates each of NB7-4-specific Dam:En and DamOnly.

Determination of NB7-4-specific Hb occupancy and chromatin accessibility, upon broad Gsb misexpression

To assay Hb occupancy and chromatin accessibility in NB7-4 under conditions of broad Gsb misexpression, we crossed 5000 females each of hsGsb;;UAS-LT3-DamOnly and hsGsb;;UAS-LT3-Dam:Hb to males of w-;19B03[AD];18F07[DBD]. Embryos were collected every hour (0-1h AEL), incubated at 25°C for 3h (3-4h AEL, Stage 7-8) and then heat shocked at 33°C just before NB delamination for 4h (7-8h AEL, Stage 11-12). Sufficient embryos were collected for 3 replicates per condition.

TaDa/CATaDa Experimental Pipeline

TaDa and CATaDa replicates were processed according to (78). Briefly, embryos were homogenized using motorized pestles, and genomic DNA was extracted using the QIAamp DNA Micro Kit (Qiagen, 56304) in a final volume of 50µL. To minimise genomic DNA shearing, wide-bore tips were used for all pipetting steps. DNA quality was assessed by loading 3 µL onto a 0.8% agarose gel. Genomic DNA was then digested with DpnI (NEB, R0176S) to cut at sites methylated by Dam or Dam:TF proteins. This was followed by DamID adaptor ligation and DpnII digestion (NEB, R0543S) to fragment unmethylated DNA. Finally, the adaptor-ligated DNA fragments were amplified by PCR (21 cycles) using MyTaq HS DNA Polymerase kit (Bioline, BIO-21112). Sequencing libraries were prepared at the Next Generation Genomics Facility (NGGF) at Bangalore Life Science Cluster (BLiSC), using the NEBNext® Ultra™ II DNA Library Prep with Sample Purification Beads (E7103L), and sequenced on the Illumina HiSeq2500 (50 base pairs single-end, 20–60 million single end reads per sample). For TaDa/CATaDa experiments involving Gsb misexpression, samples were sequenced on the NovaSeq 6000 platform (50 base pairs, 25-30 million paired-end reads per sample).

Data processing and analysis

Alignment and Filtering

Reads were aligned to the Drosophila genome (dm6) using Bowtie2 (79). The .sam files thus obtained were sorted and converted to .bam files using the SAMtools (80) and NGSUtils (81) suites. We removed reads with poor mapping quality, as well as those that aligned with blacklisted regions in dm6 or the mitochondrial genome. The resultant .bam files were used for all downstream analyses.

Pearson correlation plots were generated using the multiBam-Summary and plotCorrelation tools from the deepTools suite (82).

Peak Calling and Differential Analysis

We used MACS2 (v2.2.6) (83) to call narrow peaks on Dam:TF or DamOnly files, without controls. We used the –nolambda parameter to disable local background correction, instead calling peaks relative to a global background.

Differential analyses were performed on these peaks using DiffBind (84, 85).

(i) Determining Sites of Integration (SoIs)

SoIs were defined using Hb-TaDa data from NB5-6 and NB7-4 generated in (41). Differential analysis was performed using DiffBind, and peaks were considered differential if they met the criteria of FDR ≤ 0.01 and ≥ 2-fold change between NB5-6 and NB7-4. Shared Hb-bound sites were identified using bedtools intersect with a minimum overlap of ≥50%.

(ii) Determining genome-wide differentially accessible sites in NB5-6 and NB7-4

Genome-wide accessible and less accessible loci in NB5-6 and NB7-4 were identified by performing a differential analysis using DiffBind on Dam-bound sites in the two NBs. Peaks were considered differential if they met the criteria of FDR ≤ 0.01 and ≥ 2-fold change. To identify loci accessible in both NB5-6 and NB7-4, we used bedtools intersect with a minimum overlap of ≥50%.

(iii) Determining genome-wide STF-bound and -unbound sites

STF-bound and -unbound sites in NB5-6 and NB7-4 were determined by performing a differential analysis using DiffBind between Gsb-bound sites in NB5-6 and En-bound sites in NB7-4. eaks were considered differential if they satisfied the thresholds of FDR ≤ 0.01 and ≥ 2-fold change. Sites bound by STFs in both NBs were defined using bedtools intersect with a minimum overlap of ≥50%.

(iv) Determining genome-wide loci that are STF-only, TTF-only and STF + TTF co-bound

For this analysis, the top 5000 most confident peaks were selected for each TF— Gsb in NB5-6, En in NB7-4, and Hb in both NB5-6 and NB7-4. For each TF, these 5000 peaks included their known genomic targets. Bedtools intersect with the -v parameter was used to identify STF-only and TTF-only sites, while co-bound (STF + TTF) sites were defined using bedtools intersect with ≥90% overlap.

(v) Determining genome-wide changes in accessibility in NB7-4 upon Gsb misexpression

We used hsGsb to induce embryo-wide Gsb misexpression and assayed changes in chromatin accessibility and Hb binding in NB7-4 under these conditions. We performed a differential analysis between Dam binding in NB7-4 with and without ectopic Gsb expression. Sites were considered differential at a threshold of FDR ≤0.01 and ≥2-fold change.

Signal Correlation

We obtained signal files by using bam-Coverage from deepTools to convert all .bam files to .bigWig files. All signal files were normalized to 1× genome coverage (reads per genomic content, RPGC), thereby correcting for differences in sequencing depth between samples. All signal files were visualized using IGV Genome Browser (86, 87). To generate signal correlation plots across different peaksets, we used computeMatrix (from deepTools (82)) to obtain matrices of per-bin signal values for each replicate along a 3kb region upstream and downstream from the peak centre (bin size = 10). These matrices were imported into R, where mean signals across replicates were calculated and plotted with 95% confidence intervals. We similarly generated heat maps showing mean signals across replicates to visualize binding signal for each site.

Gsb knockdown and misexpression experiments

For knockdown experiments, crosses were set up with Vasa-Cas9;; and y[1]v[1]; gsb-gRNA. Appropriately staged embryos were collected and stained with Rabbit anti-Ap, Mouse anti-Eya and Rat anti-Ems. Stained embryos were imaged. Marker-positive cells per hemisegment were counted. The data was plotted using R.

For misexpression experiments, crosses were set up with hsGsb;; and w-;Lbe-k-Gal4,UAS-mCD8-GFP (to visualize NB5-6 lineage) or w-;;R59E09,UAS-myrGFP (to visualize NB3-5 lineage). Embryos were incubated in 18°C for 6.5h and then given a heat shock at 37°C for 30min. Subsequently, they were incubated at 23°C for 21h to stage 16 and then stained with Chicken anti-GFP, Mouse anti-Eya, Rabbit anti-Ap, and Rat anti-Ems. Marker-positive cells per hemisegment were counted. The data was plotted using R.

In all cases, box and whisker plots were generated. Statistical significance was assessed using two-tailed Mann–Whitney U tests. Significance levels are indicated as: *** p < 0.001, ** p < 0.01, * p < 0.05, ns = not significant.

Immunohistochemistry and imaging

Appropriately staged embryos were washed with distilled water and dechorionated in bleach for 2–3 minutes. After thorough rinsing with distilled water, embryos were fixed by nutating in a 1:1 mixture of 4% PFA and heptane for 20–25 minutes. Vitelline membranes were removed by vigorous shaking for 1 minute in a 1:1 mixture of heptane and methanol. Embryos were then washed in 0.5% PTX (0.5% Triton X-100 in 1X PBS), incubated overnight at 4°C with primary antibodies, washed again in 0.5% PTX, and incubated overnight at 4°C with secondary antibodies. Following additional washes in 0.5% PTX, embryos were cleared in 90% glycerol with Vectashield.

Imaging was performed at the Central Imaging and Flow Cytometry Facility (CIFF), Bangalore Life Science Cluster (BLiSC), using an Olympus FV3000 system with a 60X oil-immersion objective. Images were acquired at 512×512 or 1024×1024 resolution, viewed and manually analyzed in FIJI (ImageJ) (88), and are presented as maximum-intensity projections of z-stacks.

All figures have been assembled using Inkscape 1.4.2 (f4327f4, 2025-05-13).

Supplementary figures

Supplementary Fig. S1.

Supplementary Fig. S2.

Supplementary Fig. S3.

Supplementary Fig. S4.

Supplementary Fig. S5.

Supplementary Fig. S6.

Data availability

This study has generated DNA sequence data. The raw and processed files have been deposited to NCBI and already made public. The BioProject accession number is PRJNA1372925.

Acknowledgements

This work was supported by the Ramalingaswami Fellowship to SS, UGC-CSIR Fellowship to AB, and TIGS. We acknowledge the Fly Facility at NCBS. All sequencing was done at the NGGF sequencing facility at the Bangalore Life Science Cluster. All imaging was done at the Central Imaging and Flow Cytometry Facility (CIFF) at the Bangalore Life Science Cluster. We thank Keiko Hirono and Bhagyashree Kaduskar for their help in generating fly lines for this project. We are grateful to Vishaka Gopalan, Sridhar Hannenhalli, Sabarinathan Radhakrishnan, Bhavana Muralidha-ran, Chris Q Doe, Sachin Chanchani and Anton Iyer for invaluable discussions. We are also grateful to Chris Doe and K. VijayRaghavan, for their critical comments on the manuscript. Finally, we thank Sen Lab members for discussions and critical feedback.

Additional information

Author contributions

AB conceptualised the project, performed and analysed all experiments, wrote and reviewed the manuscript. HR con-ceptualised, performed, and analysed experiments related to the functional analysis of Gsb and reviewed the manuscript. SQS conceptualised and supervised the project and wrote and reviewed the manuscript.

Funding

Tata Institute for Genetics and Society

• Ayanthi Bhattacharya

• Sonia Q Sen

• Hemalatha Rao

Additional files

Supplementary Table S1. Summary of TaDa and CATaDa sequencing quality metrics.

Supplementary Table S2. Genomic coordinates of Sites of Integration (SoIs).

Supplementary Table S3. Genomic coordinates of differentially accessible sites in NB5-6 and NB7-4.

Supplementary Table S4. Genomic coordinates of STF-bound and -unbound sites.

Supplementary Table S5. Genomic coordinates of sites that are bound by STF only, Hb only, and co-bound by both STF and Hb in NB5-6 and NB7-4.

Supplementary Table S6. Genomic coordinates of sites that change accessibility upon Gsb misexpression.

Supplementary Table S7. Raw quantification of marker-positive cells per hemisegment (Ap, Ems, Eya, GFP-reporters) across all experimental conditions. This sheet contains the complete dataset used for all quantifications shown in Figures 7–8.