Neocortical pyramidal cells with axon-carrying basal dendrites: a prominent feature regulated by activity Authors: David Bietz, Burak Ceylan, Andrea Räk, Ina Köhler, Petra Wahle Published: 2026 ABSTRACT AbstractAxon-carrying dendrites, AcDs, have been reported in hippocampus to be privileged dendrites which can directly feed synaptic inputs into an axonal output bypassing the soma and escaping somatic inhibition. In non-primate neocortex, 15–20% of the pyramidal cells have basal AcDs and comparable proportions occurred in organotypic cortex cultures. How does an axon happen to emerge from a dendrite? Very immature migrating pyramidal neurons have no basal dendrites but already have an axon. Sprouts were seen to emerge from the very proximal portion of axons suggesting that an AcD cell starts out as a ‘dendrite-carrying axon’ (DcA) cell. Further, the formation of basal AcD was influenced by neuronal activity. Inhibition of GluN2B receptors, overexpression of strongly calcium binding indicator proteins and activity deprivation impaired basal dendritic growth and reduced the proportion of AcD pyramidal neurons. In contrast, perinatal enhancement of glutamatergic network activity with kainate or carbachol via subplate neurons accelerated the appearance of AcD pyramidal neurons. None of the treatments altered the proportion of AcD interneurons. Together, the larvae stage of a cortical AcD pyramidal neuron can well be the DcA configuration, and the formation of basal AcDs was - at least transiently – regulated by neuronal activity, calcium and glutamatergic signaling.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1007/s00429-026-03180-y.Keywords: Activity-dependent neurite growth, NMDA receptor GluN2B and GluN2D, Calcium, Basket cells, GABA-ergic inhibitory neurons, Organotypic cortex cultures Introduction Neurons have several dendrites and one special neurite, the axon. Since the times of Ramon y Cajal (1911) typical morphological criteria distinguish the axon from dendrites, all confirmed decades later by neurite-specific molecular signatures. The majority of human neurons in Cajal’s drawings have axons arising from the cell body. Axons start with a smooth, tapering hillock which is followed by the axon initial segment (AIS), the electrogenic domain generating the action potential (Kole and Brette 2018). The axon continues as non-tapering descending fiber which in pyramidal cells gives rise—often at right angles or recurrent—to thinner collaterals with varicosities. However, axons may as well emerge from a dendrite, a feature now referred to as ‘axon-carrying dendrite’ (AcD) and ‘AcD neurons’. Functionally, mouse hippocampal CA1 AcD pyramidal cells are privileged since the input onto the AcD bypasses the soma, circumvents somatic integration and somatic inhibition, and can initiate action potentials in the axon (Thome et al. 2014). AcDs are intrinsically more excitable, generating dendritic spikes with higher probability and greater strength. Synaptic input onto AcDs efficiently elicits action potentials with lower thresholds compared to those of non-AcDs due to the short electrotonic distance between input and AIS. These features enable allocortical CA1 AcD pyramidal cells to generate ripple oscillations (Hodapp et al. 2024). Whether or not the same accounts for AcD pyramidal cells of the neocortex and how the AcD configuration affects the working mode of the inhibitory interneurons remains to be shown. AcDs occur in pyramidal neurons of all neocortical layers in rodents, lagomorphs, ungulates, and carnivores, infant and adult macaque and adult and aged human (Peters et al. 1968; Smit and Uyling 1975; Van der Loos 1976Peters and Kara 1985a; Ferrer et al. 1986a, b; Hübener et al. 1990; Reblet et al. 1992; Matsubara et al. 1996; Prieto and Winer 1999; Mendizabal-Zubiaga et al. 2007; Hamada et al. 2016; Ernst et al. 2018; Wahle et al. 2022). In non-primate species, 15–20% of pyramidal cells of layers 2–6 have basal AcD. Axons from apical dendrites are extremely rare. The proportion of basal AcD was lower in macaque and particularly low for supragranular neurons of the human neocortex. Ontogenetically older pyramidal cells of subplate and interstitial cells of the white matter exhibit AcDs to much higher percentages than gray matter pyramidal cells suggestive of evolutionary regulation and cell-type specificity. For instance, subplate axonal loop cells rarely have axons from dendrites (Wahle et al. 2022). Gray matter interneurons and Cajal-Retzius cells have AcDs at higher percentages than pyramidal cells (Retzius 1893; Jones 1975; Meyer 1982; Meyer and Wahle 1988; Kisvárday et al. 1990; Wahle 1993; Höfflin et al. 2017; Wahle et al. 2022; Sobierajski et al. 2025). The proportion of AcD pyramidal cells in mouse CA1 ranges from about 20% (Benavides-Piccione et al. 2020) determined with dye injections to about 50% (Thome et al. 2014) with genetic labeling. Differences in proportions could have been due to sampling strategies since the majority of rodent CA1 AcD pyramidal cells resides in the superficial sublayer of the central part of CA1 in dorsal hippocampus, with much lower proportions in ventral hippocampus and subiculum (Thome et al. 2014). Alternatively, differences of methods could contribute, such as the validation of the AIS done in the latter study to unequivocally identify the axon in > 700 neurons. In human CA1, AcD pyramidal cells represent 44% (Benavides-Piccione et al. 2020). A variance by a factor of two has also been found in Thy-1 labeled mouse neocortical pyramidal cells (Wahle et al. 2022) either due to individual differences of Thy-1 expression or to biological variability. This indicates that only the assessment of large numbers of neurons will yield reliable data. A recent study (Han et al. 2025) in dissociated hippocampal neurons of Wistar Unilever HsdCpb: WU (Envigo) rats reports AcD neurons as early as 3 days after plating before the onset of synaptogenesis. Interestingly, from days in vitro (DIV) 3 to DIV 5/7 the proportion of hippocampal AcD neurons increases to up to 20% followed by a decrease to about 10% at DIV 21 (Han et al. 2025). The in vitro proportions are lower than proportions reported in CA1/CA3 in vivo (Thome et al. 2014; Hodapp et al. 2022) but comparable to proportions reported in adult non-primate neocortex (Wahle et al. 2022). Importantly, Han et al. (2025) have meticulously worked out that the axon grows first, that the dendrite which lateron will carry the axon emerges as collateral of the axon close to the soma, and that the joint root harbors tyrosinated microtubuli in an orientation typical for axons. In other words, the ‘axon-is-first’ observation in dissociated cells (Han et al. 2025) could suggest that an AcD cell actually starts as a ‘dendrite-carrying axon’ (DcA) cell. Here, we assessed rat visual cortical neurons in organotypic slice cultures which preserve the 3D environment and laminar organization with apical-basal cellular polarity as well as becoming spontaneously active. From their earliest stage, cortical pyramidal neurons extend an axon from the basal cell pole. It is the primary axon, and not a secondary one forming e.g. after plating dissociated cells. Basal dendrites sprout once the migrating cells reach their laminar destination. A future AcD may well happen to sprout from the axon close to the soma as reported (Han et al. 2025). Alternatively, during basal dendritogenesis the existing axon could hitchhike a dendrite which by chance happened to sprout near the axon’s hillock. The ‘shared root’ configuration (Wahle et al. 2022) could be an intermediate state. Since dendrite development is strongly activity-dependent we analyzed if early postnatal network activity could influence the proportion of AcD pyramidal cells and interneurons. The results suggested that the basal AcD configuration of neocortical pyramidal cells is a feature regulated by spontaneous activity. Materials and methods Organotypic cultures The slice culture material was mostly archived from previously published studies supplemented with fresh transfections mainly to address the basket cells (BC with local plexuses or horizontal projections) and the subset of non-basket cells (non-BC with translaminar vertically projecting bitufted cells and cells with arcade axons). ‘Roller-type’ organotypic slice cultures (OTCs) were prepared from P0-1 newborn Long-Evans hooded rat visual cortex using 4–6 pups of one litter per batch and cortex slices of each pup were allocated to every condition tested. Biolistic transfection and histology were done as described (Jack et al. 2019; Gonda et al. 2020; Gasterstädt et al. 2020). Briefly, plasmids were coated to 1 μm gold particles blasted with a Helios gene gun (BioRad, Munich, Germany) with 180 psi helium pressure onto the cultures. The particles non-selectively penetrate cellular nuclei, and cells begin to transcribe the plasmid DNA. The fairly small reporter protein(s) diffuse throughout the cytosol resulting within 2–3 days of expression in a complete filling of the cell’s dendrites and ideally the complete axons (see Fig. 1). We used plasmids encoding EGFP or mCherry under CMV promoter (pEGFP-N1 in pcDNA3.0, Clontech, Heidelberg, Germany, cat. no. 632370; pCMV-mCherry-N1, Clontech, Heidelberg, Germany, cat no. 632523) or, in the recent experiments plasmids encoding EGFP under the Dlx enhancer (Fishell-GFP; pAAV-mDlx-Fishell-1, addgene, cat. no. 83900) which expresses intensely in interneurons besides pyramidal cells and thus increased the number of interneurons classifiable as BC and non-BC. Slices were fixed in the culture tube with prewarmed (25–30 °C) buffered 4% paraformaldehyde with 0.1% glutardialdehyde added directly to the medium for 10 min, and 4% paraformaldehyde for another 30 min, 3 × 5 min rinses with phosphate buffer pH 7.4, permeabilization with 0.3% Triton X-100 in phosphate buffer, and incubation with 5% bovine serum albumin in TBS for 3–4 h. Immunostaining (mouse anti-EGFP, Merck, clone GSN24, RRID: AB_563117, 1:1000; mouse anti-mCherry, TakaraBio, cat. no. 632543, RRID: AB_2307319, 1:1000; incubated for 30–48 h) was followed by biotinylated secondaries (goat anti-mouse, Dako A/S, cat. no. E0433, RRID: AB_2687905, used 1:750; incubated for about 12 h), and detection with ABC reagent (Vector Laboratories, PK-7100, RRID: AB_2336827, incubated for 3–4 h) and OsO4 intensification of the horseradish peroxidase-diaminobenzidine reaction product. Pyramidal neurons with axons from soma or basal dendrites. (a) Somatic axon from the basal pole opposite to the apical dendrite. Dendrites were spiny, the axon was smooth. (b) Axon emerging from a lateral aspect of the soma. (c) Inverted pyramidal cell with axon taking a hairpin turn to descend to the white matter. (d) Basal AcD cell, photomontage to show the descending axon; upper inset, the axon origin at higher magnification; lower inset, overexposure shows that the axon and the dendrite merge. (e) Pyramidal cell suggestive of a ‘shared root’ configuration; inset, overexposure reveals the axon was arising from the dendrite. f, g, h Axons arising from basal dendrites at variable distance to the soma. i Axon arising from a thick horizontal basal dendrite taking a turn down to the white matter. Cells sampled from DIV 10–20 control cultures and arranged with the former pial surface to the top. Arrows indicate axons, asterisks mark axonal branch points. Isolated black grains: gold particles. Scale bars: 20 μm Experimental design For the NMDA receptor study (Gonda et al. 2020) ifenprodil and Ro25-6981 (both from Sigma-Aldrich) have been used to antagonize GluN2B-containing NMDA receptors. To inhibit GluN2A, NVP-AAM077 (from Sigma-Aldrich) and TCN201 (from Tocris) were used. Ro25 and TCN201 were dissolved in dimethylsulfoxide and applied at 5 µM and 10 µM (f. c.) in the medium, respectively. Controls were mock-stimulated with the same final concentration of dimethylsulfoxide. The drugs were supplemented to the culture medium from DIV 5–10, and DIV 15–20 for analysis at DIV 10 and DIV 20, respectively. For the fresh experiments (prepared during the last months to increase the n of batches) we used only ifenprodil (stock dissolved in ethanol and further diluted with water; applied at 10 µM f. c. in the culture medium) and NVP-AAM077 (stock dissolved in water; applied at 200 nM f. c. in the culture medium). Controls received water. In addition, we did a DIV 5–10 treatment with maintenance in normal medium until DIV 15. Additionally, we assessed slice cultures treated with CIQ and DQP to modulate GluN2D-containing NMDA receptors (Köhler et al. 2025) expressed by basket and non-basket cells. The activator CIQ (3-chlorophenyl)(6,7-dimethoxy-1-((4-methoxyphenoxy)-methyl)−3,4-dihydroisoquinolin-2(1 H)-yl)methanone) and the inhibitor DQP (4-(5-(4-bromophenyl)−3-(6-methyl-2-oxo-4-phenyl-1,2-dihydroquinolin-3-yl)−4,5-dihydro-1 H-pyrazol-1-yl)−4-oxobutanoic acid) (both from Tocris) were dissolved in dimethylsulfoxide. Drugs were applied daily at 15 µM f. c. in the medium from DIV 5–10 or DIV 15–20 for analysis at DIV 10 and DIV 20, respectively. Controls were mock-stimulated with the same final concentration of dimethylsulfoxide. For the genetically encoded calcium indicator (GECI) study (Gasterstädt et al. 2020), OTCs were transfected at DIV 3–4 with CMV-driven plasmids encoding GCaMP3 or TN-XXL (G3/TN, data pooled) as well as GCaMP6m or GCaMP6m-Xc (G6m, data pooled) with analysis at DIV 14 and DIV 22–25. For the chronic activity deprivation, unpublished slice culture material prepared during the past years was reassessed and supplemented with freshly prepared culture batches. Culture medium has been supplemented with 10 mM (f. c.) MgSO4 from DIV 2 onwards. Control OTCs received medium with 2 mM MgSO4. Besides assessing AcD proportions we reconstructed pyramidal cell dendrites to determine length and number of segments at DIV 10 and DIV 20. For the the kainate (KA) study (Jack et al. 2019), 500 nM KA (f. c.; from Tocris) has been added to the culture medium from DIV 5–10 with analysis at DIV 10. Control cultures received water. We repeated this design to increase the number of addressable interneurons and added an early stimulation from DIV 1–5 with assessment at DIV 5, and 15–20. For the carbachol (CCH) study (Wahle et al. 2026), 1 µM (f. c.; from Sigma-Aldrich) CCH has been added to the culture medium daily from DIV 1–5 followed by maintenance in normal medium. Control cultures received water. We repeated this design to increase the number of addressable pyramidal cells and interneurons. Rationale, major findings and time line of the experiments are summarized at the beginning of the Results chapters and the Figures. Definition of AcDs and cell types addressed Using conservative criteria (Wahle et al. 2022) we scored an AcD pyramidal cell when a gap of minimum the width of an axon hillock exists between the hillock and the somatic wall, and/or when an idealized vector through the axon hillock bypasses the soma. Axons emerging from the broad spine-less base of an apical pyramidal cell dendrite were scored as somatic axons, as was the ‘shared root-configuration’. The electrophysiological behavior of ‘shared root’ cells is not different from non-AcD cells (Han et al. 2025). In rat, pyramidal cells of upper and lower layers display the same proportion of AcD pyramidal cells (Wahle et al. 2022). Therefore, cells were sampled throughout the depth of the OTC, irrespective of laminar position, to achieve the numbers required for decent statistics in slice culture material prepared with sparse transfections. Subplate/L6b had somewhat higher proportions of AcD neurons (Wahle et al. 2022) but represents a minor compartment compared to the L2-6a in rodent. The few cells which might have happened to be sampled in the one or the other OTC would have not substantially altered the AcD cell proportions. The majority of the neurons assessed here were sampled from supragranular layers. Spiny cells without clear apical-basal dendritic polarity (stellates) were omitted. Axons descending towards the white matter were sampled when seen beyond the first 1–2 recurrent or horizontal collaterals. Axons from basal AcD emerged mostly within the proximal or a secondary segment. In OTC, we regularly encountered axons emerging from pyramidal cell apical dendrites. Some emerged at tens of micrometers distant from the soma usually U-turning down to the white matter, a feature resembling avian tectal Shepherd’s crook cells (Domesick and Morest 1977; Woodson et al. 1991; Lischka et al. 2018). We cannot exclude having overloocked far distant apical axon origins. However, axons more than 100 μm away from the soma were seldom encountered, and we assume it is not a major fraction which was potentially underestimated. In vivo, AcDs are basal dendrites, only extremely few apical AcDs are seen (Wahle et al. 2022). We included interneurons with clearly recognizable type-specific axonal features adressable as BC and non-BC (Kaiser et al. 2001; Gonda et al. 2023; Köhler et al. 2025; Wahle et al. 2026). Martinotti cells were included in the present set of non-BC; all are dendrite-targeting interneurons. Bipolar cells were excluded, their small somata were too rarely transfected, and most bipolar cells are AcD cells (Peters and Kara 1985b). Data sampling, data presentation and statistical analysis Counting was done by two trained observers initially with the 4-eye principle (Zeiss discussion bridge) at 1000x magnification, lateron with single observers who were blinded to condition. For pyramidal cells, we set a minimum of 20 neurons per data point. In well transfected OTCs where a minumum of 20 pyramidal cells was retrieved, the OTC-average was computed and used to plot. In sparsely transfected batches, values from up to 8 OTCs were pooled to generate one data point, always taking care to have about equal numbers from the batch-internal control cultures. In rodent, interneurons represent only about 20% of the cortical neurons, and thus were sparsely occuring in the slice culture material. We set a minimum of 10 BC and 10 non-BC per data point. For nearly all batches, all BC and all non-BC sampled had to be pooled such that every culture batch yielded only one data point. In case the minimum was not reached, we pooled across batches. Proportions of AcD cells obtained in the archived slice culture material and the freshly transfected batches were fully comparable. Neither the type of plasmid nor the staining reagents, the culture media or the rat breeding population used over the many years had detectable influences on the outcome. Supplemental Table S1 and Supplemental Table S2 summarizes the number of batches, of OTC and of neurons assessed. Fotomicrographs were taken at 200x, 400x, and 1000x magnification with a Zeiss Axiophot equipped with color CCD camera and DIC optics. Neurites meander in thick cultures and portions of the neurites were not always exactly in the focal plane. Completeness would have required montages (done in Fig. 1d for the axon). For all remaining fotographs, we decided against in order to avoid any picture manipulation of micron-sized objects. Contrast enhancement and color adjustment was done with Photoshop. Pictures were arranged and labeled with Powerpoint. Data are presented as box plots (median, 25%/75% quartiles, 5%/95% whiskers) and/or for the interneuron classes additionally as dot plots. Further, we plotted the batch averages of control and treatment groups for selected stages to confirm that statistic holds even when reducing the number of values. Graphics and statistical analysis were done with Sigma Plot 12 for Windows (SPSS Incorporated, Chicago, IL, USA). Non-parametric Mann–Whitney U-tests or ANOVA-on-ranks (Dunn’s correction for multiple tests) versus control were employed to assess significant differences between treatment groups and control. Results Axons from dendrites and axons from the soma Pyramidal neurons had somatic axons (Fig. 1a–c) or axons from a basal dendrite (Fig. 1d–i). Careful microscopy was required and turning up the light helped to determine the point of axon origin (insets Fig. 1d, e). Somatic axons and AcD also occurred in BC (Supplemental Fig. S1a–c) and non-BC (Supplemental Fig. S1d–f). The proportion of AcD pyramidal cells increases during development We compiled the proportion of AcD pyramidal cells and AcD interneurons from control OTCs of numerous batches prepared over the past 12 years including non-related experiments, and also the freshly prepared batches. The proportion of basal AcD pyramidal cells was low at DIV 5 (Fig. 2a). This was expected because numerous neurons were still migrating and had not yet formed any basal dendrites. The proportion increased at DIV 10 followed by a plateau until DIV 30. In visual cortex, pyramidal cells are by and large mature around P21 with respect to spine density (Miller 1981), although some dendritic remodeling proceeds through the second postnatal month (Romand et al. 2011). A developmental increase from ~ 30% AcD pyramidal cells at postnatal day 8 to about 50% at postnatal days 28–25 has also been reported in mouse CA1 (Thome et al. 2014). Developmental aspects. a The proportion of pyramidal cells [in percent from total pyramidal cells counted] with basal AcD increased from DIV 5–10 followed by a plateau. Mann-Whitney test, p value is indicated. b The proportion of AcD basket cells was unchanged during development. c The proportion of AcD non-basket cells was unchanged during development. DIV, days in vitro. The number of batches, of OTC and of neurons assessed is given in Supplemental Table S1 In rat visual cortex, symmetric axosomatic synapses are present at postnatal day 6 (Miller and Peters 1981). Interneuron types become recognizable around postnatal day 10 in vivo (Micheva et al. 2021) and DIV 10 in OTCs (Köhler et al. 2025; Wahle et al. 2026). Interneurons represent about 20% of the cortical neurons. Therefore, they were less often transfected even with Fishell-GFP labeling, and of those not all were well enough stained to be classified by axonal morphology. Thus, interneurons were sampled in much smaller numbers. In BC and non-BC, the proportions of AcD neurons were similar with 40–50% being AcD cells. Further, already at DIV 10, the earliest stage assessed, proportions were similar to those seen in older OTCs (Fig. 2b, c). Of note, the batch-to-batch variability was substantial resulting in larger boxes and whiskers which repeats in the following experiments. Given that the proportions of BC and non-BC AcD cells were so similar we pooled the interneurons for documenting the effects of the experimental treatments on interneuronal AcD proportions. Data separated for BC and non-BC are given in Supplemental Fig. S2a-j. AcD pyramidal cell development In dissociated hippocampal neurons Han et al. (2025) have imaged the molecular identity and function of the AcDs and the AIS with multi-labeling and tracing approaches. Han et al. (2025) have not specifcally adressed the interneuron types likely because they were not yet fully present or recognizable in dissociated cultures. Further, the type of dendrite, apical or basal, is hard to distinguish in dissociated neurons. These questions can be ideally addressed in OTCs. Most remarkable is the ‘axon-is-first’ observation (Han et al. 2025). For cortical pyramidal cells this is necessarily so because migrating cells have an axon from the earliest time point of their inside-out journey but not yet basal dendrites (Miller, 1981). At DIV 5, the earliest time point assessed, immature presumptive pyramidal cells had very small somata, an axon descending from the ventral pole towards the white matter (Fig. 3a, b), a leading process towards the pial surface but no basal dendrites. More advanced pyramidal cells had a more branched apical tree, and short basal dendrites, and some of them already carry the axon with the first collaterals at on average 30–60 μm distance to the soma. Some immature pyramidal cells had the typical somatic axon, and its very proximal portion displayed a short excrescence (Fig. 3c-e). In other such cells, the short sprouts became fuzzy, appeared to elongate (Fig. 3f-h), start to branch in a pattern typical for dendrites (Fig. 3i), and form filopodia and spines until, finally, a basal AcD has emerged (Fig. 3j, k). At DIV 10 and older, axons arose with smooth hillocks and proximal excrescences were no longer observed (Fig. 1). This DIV 5 series of pictures suggested that a dendrite can emerge from the very proximal portion of the axon, and consequently that the DcA configuration could be the earliest form of a basal AcD at least in some pyramidal cells. The axon origin of immature pyramidal neurons. a, b Presumptive pyramidal cells with axon from the basal pole, a smooth hillock, no collaterals. c, d, e Presumptive pyramidal cells with conspicuous excresences at the axonal origin very close to the soma; the axons have formed collaterals at some distance to the soma. f, g, h. Excresences unfolding into outgrowing sprouts. Note that the first axonal collateral branched off at a much larger distance. i Short processes at the axonal origin. j The axon emerged from a short outgrowing dendrite. k A branching dendrite gives rise to an axon. All cells sampled from supragranular layers from the control cultures of one batch transfected at DIV 3 and stained at DIV 5. The b/w insets show the cells and axonal trajectories. Arrows indicate axons, asterisks mark branch points. Arrowheads indicate excrescences and sprouts at the axon’s origin. Scale bars: 20 μm Inhibition of GluN2B receptors reduced the proportion of AcD pyramidal cells Strongly calcium-mobilizing GluN2B receptors are richly expressed in early postnatal cortex and known to regulate dendritic growth (Rajan and Cline 1998). Antagonizing NMDARs from P1–P21 reduces the length of basal but not apical dendrites of rat prefrontal cortical pyramidal cells (Wedzony et al. 2005). Inhibiting GluN2B-containing receptors until DIV 10 in OTC substantially impairs basal dendritic growth (Gonda et al. 2020). Inhibiting GluN2A until DIV 10 had no such effect. GluN1/GluN2B receptors decrease during development in vivo and in OTC due to the gradual upregulation of GluN2A and the emergence of heterotrimeric NMDA receptors. Indeed, antagonizing GluN2B-containing receptors at DIV 15–20 no longer altered dendritic growth. This has argued for a period up to DIV 15 during which immature basal dendrites require calcium influx through NMDA receptors for proper growth (Gonda et al. 2020). Thus, the hypothesis is that blocking GluN2B signaling and calcium influx in an early time period will result in lower proportions of AcD pyramidal neurons. Treatment was done from DIV 5–10 with analysis at DIV 10, DIV 5–10 with analysis at DIV 15 and DIV 15–20 with analysis at DIV 20 (Fig. 4a). Treatment with ifenprodil/Ro25 resulted in lower proportions of AcD pyramidal cells at DIV 10 and DIV 15 (Fig. 4b), also when looking at the batch averages (Fig. 4c). Blocking GluN2A receptors did not change the proportions, and served as batch-internal control (Fig. 4b). During the late time window, neither inhibiting GluN2B nor GluN2A changed the proportions. This suggested that the emergence of AcD is related to the formation of basal dendrites, and further suggested the existence of a critical period during which AcD formation is sensitive to activity. Proportions of AcD interneurons remained at control level (Fig. 4d). Proportion of pyramidal cells with basal AcD was reduced by inhibiting GluN2B-containing NMDA receptors. a Time line of pharmacological treatment with analysis at DIV 10, 15, and 20. b Box plots revealed the reduction after early inhibition but no longer after exposure from DIV 15–20. Inhibiting GluN2A did not alter the proportion. c Batch averages plotted versus the batch-internal control (in the middle) at DIV 10 and DIV 15 for Ifenprodil/Ro and NVP/TCN treatment. Every culture batch is represented by dots connected by lines. d The proportion of AcD interneurons was not altered. Symbols in this and the following figures: flash, time of biolistic transfection; colored downpointing arrow, treatment; upward arrows, time of fixation with staining and analysis. ANOVA on ranks test versus control in b, c, asterisk marks the group significantly different from control, p values are given. The number of batches, of cultures, and of neurons assessed is given in Supplemental Table S2 In addition, interneurons were directly activated with CIQ, an agonist at GluN2D-containing NMDA receptors or inhibited with the GluN2D antagonist DQP. DQP treatment subtly increases the network activity via disinhibiting pyramidal cells leading to an increase of apical dendritic growth and higher spine densities without changing the complexity of basal dendrites at DIV 10 and DIV 20 (Köhler et al. 2025). As expected, there was no change in the proportion of pyramidal cells with basal AcD (Supplemental Fig. S3a-c). This was in line with the suggestion that the emergence of AcD is related to basal dendritic growth. Somewhat unexpected, there was also no change in the proportion of AcD BC and AcD non-BC (Supplemental Fig. S3d, e). Strong buffering of intracellular calcium reduced the proportion of AcD pyramidal cells By their very nature, GECI’s bind and buffer calcium (McMahon and Jackson 2018). Our study has revealed that overexpression of ‘older’ (first-generation) G3/TN strongly impair basal dendritic growth at DIV 14 and DIV 22–25, whereas overexpression of more recently developed GCaMP6 variants (G6m) seems much more ‘neuron-friendly’ (Gasterstädt et al. 2020). The hypothesis is that blocking calcium influx in an early time period will lead to lower proportions of basal AcD pyramidal neurons. Expression was from DIV 3–4 with analysis at DIV 14 and at DIV 22–25 (Fig. 5a). Indeed, the proportion of basal AcD of G3/TN expressing pyramidal cells was substantially lower at DIV 14 (Fig. 5b), also when looking only at the batch averages (Fig. 5c). Proportions were at control level at DIV 22–25 arguing for h a substantial delay of AcD development. G6m expressing pyramidal cells had proportions of AcD neurons similar to control at DIV 14 and DIV 22–25 (Fig. 5b). This supported the above suggestion of a dependence on basal dendritic growth and of the existence of an early critical period during which AcD formation is sensitive to calcium signaling. Proportions of AcD interneurons remained at control level (Fig. 5d). Proportion of pyramidal cells with basal AcD was reduced by overexpression of GECIs. a Time line of expression from DIV 3–4 with analysis at DIV 14 and at DIV 22–25. b Box plots revealed the reduction after early expression of G3/TN but not G6m at DIV 14. c Batch averages plotted versus the batch-internal control (in the middle) at DIV 14 for G3/TN and G6M transfections. Every culture batch is represented by dots connected by lines. d. The proportion of AcD interneurons was not altered. ANOVA on ranks test versus control in b, c, asterisk marks the group significantly different from control, p values are given. The number of batches, of cultures, and of neurons assessed is given in Supplemental Table S2 Activity deprivation reduced the proportion of AcD pyramidal cells With 10 mM MgSO4, calcium influx through voltage-gated channels as well as action potential firing of neurons in OTC are suppressed and the expression of activity-dependent genes is low (Gorba et al. 1999). For instance, BDNF mRNA remains low and BDNF is a major driver of dendritic growth of pyramidal cells and interneurons which heavily express TrkB receptors (Gorba and Wahle 1999; Polleux et al. 2002; Wirth et al. 2003). The hypothesis is that blocking spontaneous action potential firing and calcium influx in an early time period will lead to lower proportions of AcD pyramidal neurons. Treatment with 10 mM MgSO4 was started at DIV 2 with analysis at the ages indicated; transfections were done to allow at least 3 days of EGFP or Fishell-GFP expression (Fig. 6a). At DIV 15–20, the proportion of AcD pyramidal cells remained at the level seen at DIV 10 (Fig. 6b), also when looking the batch averages (Fig. 6c) suggestive of a substantial delay of AcD development. Only at DIV 30 had the proportion recovered to the age-matched control levels. Proportions of AcD interneurons remained at control level (Fig. 6d). The GluN2B and the G3/TN data report a reduction of basal dendritic growth at DIV 10 (Gonda et al. 2020) and at DIV 14 and 22–25 (Gasterstädt et al. 2020) which was now related to lower proportions of AcD pyramidal cells. We reconstructed pyramidal cells (> 80% from supragranular layers) from the MgSO4 exposed OTCs. Indeed, we observed shorter and less branched basal dendrites at DIV 20 but not yet at DIV 10 (Supplemental Table S3). Interestingly, length and branching of the apical dendrites was not depressed by MgSO4 exposure. The result supported the suggestion that the basal dendritic AcD configuration is linked to basal dendritic differentiation. Moreover, it suggested that basal dendrites can initially grow without requiring action potential activity. Beyond DIV 10 however, basal dendrites remained hypomorphic in the absence of action potential activity. Proportion of pyramidal cells with basal AcD was reduced by activity deprivation. a Time line of deprivation starting from DIV 2 with analysis at DIV 5, 10, 15–20, and 30. b Box plots revealed the reduction at DIV 15–20. c Batch averages plotted versus the batch-internal control at DIV 15–20. Every pair of dots represents one culture batch. d The proportion of AcD interneurons was not altered by activity deprivation. Rank sum tests in b, c, p values are given. The number of batches, of cultures, and of neurons assessed is given in Supplemental Table S2 Kainate-evoked network activity increased the proportion of AcD pyramidal cells At nanomolar doses, KA preferentially activates metabotropic receptors which in early cortex (Beed et al. 2009; Lauri and Taira 2011; Valbuena and Lerma 2016) elicits network activity which robustly enhances dendritic growth (Jack et al. 2019). The hypothesis is that enhancing spontaneous activity in an early time period will lead to higher proportions of AcD pyramidal neurons. KA was applied once daily with analysis at DIV 5, 10 and 15–20 (ages pooled) (Fig. 7a). The developmental graph (Fig. 2a) already revealed a substantial proportion of AcD pyramidal cells at DIV 5. KA-mediate network activity increased this proportion substantially at DIV 5 (Fig. 7b), also when looking at the batch averages (Fig. 7c). Proportions of AcD interneurons remained at control level (Fig. 7d). Proportion of pyramidal cells with basal AcD was increased by kainate-mediated early network activity. a Time line of stimulation in two early time windows with analysis at DIV 5, 10, and 15–20. b. Box plots revealed the increase at DIV 5 (significant) and at DIV 10 (statistical trend, p = 0.06). c Batch averages of KA-stimulated cultures versus the batch-internal control at DIV 5 and DIV 10. Every pair of dots represents one culture batch. d The proportion of AcD interneurons was not altered by KA stimulation. Rank sum tests in b, c, p values are given. The number of batches, of cultures, and of neurons assessed is given in Supplemental Table S2 Network activity via mAChR activation of subplate/L6b neurons increased the proportion of AcD pyramidal cells Muscarinic M1/5 acetylcholine receptors are richly expressed in perinatal neocortex by glutamatergic subplate neurons with projections into the gray matter. This way, the mAChR agonist CCH enhances glutamatergic network activity in ex vivo neonatal mouse neocortex (Dupont et al. 2006) as well as in OTCs. Subsequently, this network activity promotes BDNF mRNA expression, RAS-ERK signaling, enhances expression of synaptic and interneuronal proteins, and promotes dendritic and axonal growth of pyramidal cells and interneurons (Wahle et al. 2026). The hypothesis is that enhancing spontaneous activity in an early time period will lead to higher proportions of AcD pyramidal neurons. CCH was applied once daily from DIV 1–5 with analysis at DIV 5, 10, 15, and 20 (Fig. 8a). Indeed, CCH treatment evoked substantial larger proportions of AcD pyramidal cells at DIV 5 and DIV 10 (Fig. 8b), also when looking at the batch averages (Fig. 8c). Proportions of AcD interneurons remained at control level (Fig. 8d). Proportion of pyramidal cells with basal AcD was increased by mAChR-mediated early network activity. a Time line of CCH stimulation with analysis at DIV 5, 10, 15 and 20. c Box plots revealed the increase at DIV 5 which persisted at DIV 10. c. Batch averages of carbachol-stimulated cultures versus the batch-internal control at DIV 5 and DIV 10. Every pair of dots represents one culture batch. d The proportion of AcD interneurons was not altered. Rank sum tests in b, c p values are given. The number of batches, of cultures, and of neurons assessed is given in Supplemental Table S2 The proportion of AcD basket and non-basket cells was neither regulated during development nor influenced by activity Interneurons were surprisingly unresponsive to depressing or enhancing activity. Analysis of BC (Supplemental Fig. S2a-e) and non-BC (Supplemental Fig. S2f-j) separately also did not reveal any systematic response to the treatments. Not even the direct modulation of interneuronal activity with CIQ or DQP changed the proportions (Supplemental Fig. S3d, e). Data points in some of the currently tested experimental conditions were still few, and the data scatter substantially due to the high OTC-to-OTC and batch-to-batch variability. The statement could be seen as preliminary. However, due to sampling numerically small subsets with our approach will make it very difficult to come to other conclusions. The proportion of apical dendritic AcD pyramidal cells was regulated during development but not by the experimental conditions tested In contrast to neocortex in vivo (Wahle et al. 2022), we regularly encountered axons arising from apical dendrites (Supplemental Fig. S4a-f). Careful microscopy e.g. with light overexposure is necessary to make a correct decision (Supplemental Fig. S4a). Cells with somatic axons and apical AcD resided next to each other (Supplemental Fig. S4b) and next to cells with basal AcDs. Axons from apical AcDs were seen as early as DIV 5 (Supplemental Fig. S4a-c) and at older ages (Supplemental Fig. S4d-f). They always started with a smooth hillock without excrescences. The proportion of apical AcD pyramidal cells increased during development in OTCs (Supplemental Fig. S5a). However, the proportions did not change in any systematic way with the experimental conditions tested (Supplemental Fig. S5b-f). Discussion The ‘axon-is-first’ observation (Han et al. 2025) has led to the view that at least a fraction of basal AcD starts from the proximal portion of the axon to the consequence that such a DcA configuration is the larvae stage of an AcD pyramidal cell. When preparing OTCs at P1-2, cortical neurons destined for supragranular layers are outwardly migrating with their leading process becoming the apical dendrite, and basal dendrites not yet formed (Miller 1981). All have an axon towards the white matter. When chopping 350–400 μm thick slices, neurons within the core of the slice preserve their primary axons. We suggest it is this fraction of cells which preferentially bears the proximal axonal excrescences which elaborate into a dendrite. It cannot be axonal branches simply because axons collateralize only distal to the AIS and many of the cells examined (Fig. 3) did already have collaterals at an appropriate distance to the soma. Axons import neurotrophic factors, paracrine or from distal synaptic partners. A proximal excrescences can use the factors to rapidly grow into a full dendrite and the axonal hillock would hitchhike rather passively “away” from the soma. This way, axons from basal AcD may reside at substantial distance to the soma. Mechanical stress could also play a role. In OTCs as in vivo, axons quickly ramify, interact with the ECM and presynapses stick to target cells with adhesion molecules. In case the soma happens to translocate during flattening of the OTC an elongating AcD might serve as a bridge that keeps the axon connected despite of mechanical tension. Reliable differences in the proportions became evident with assessing a high number of batches and neurons. In fact, proportions of neocortical AcD pyramidal cells varied from OTC to OTC and batch to batch. This might have been a consequence of the relatively small numbers of addressable transfectants per culture given that the archived slice culture material intentionally had low transfection rates. Variability of proportions could be further due to heterochronous maturation. Pups of a given litter vary in body and brain weight, larger litters are usually born a bit earlier, having lower body weights which can result in a developmental delay of e.g. the thalamocortical projection (Hoerder-Suabedissen et al. 2008). Moreover, not all neurons destined for a given layer are generated at the same moment, nor arriving at their laminar destination or differentiating at the same time. Basal dendrites form over a protracted period of time, and proportions of basal AcD pyramidal cells increased up to DIV 10/15 in OTCs. This is evident from Fig. 3 with cells sampled from one batch. Technical reasons can contribute. In vivo, axons from apical dendrites are extremely rare (Wahle et al. 2022). Could the high proportion of pyramidal cells with apical AcD observed in our study argue for an abberant differentiation? OTCs preserve the laminar environment and cellular polarity. Unavoidably however, during preparation primary axons at the cut sides of the slice became severed. Furthermore, handling stress and ongoing flattening of the OTCs might inflict mechanical load which could disrupt long neurites forcing the neurons the replace a severed axon. If the axonal stump remains long enough, the primary axon is known to regenerate from the stump. When disrupting the axon closer to the soma for instance of immature hippocampal cells, dissociated and in OTCs, a majority of neurons will sprout a new axon from a different neurite (Goslin and Banker 1991; Rodriguez et al. 2001; Gomis-Rüth et al. 2008). For neocortical neurons with basal dendrites not yet present, the maturing apical dendrite may be able to form a secondary axon quicker than the soma can do. Interestingly, the DQP treatment promotes apical dendritic growth via disinhibition (Köhler et al. 2025). Yet, this did not influence the proportion of pyramidal cells with apical AcD suggesting that the appearance of apical AcD is not related to dendritic growth. Tectal Shepherd’s crook cells form the primary axon far distally at the ascending dendrite (Domesick and Morest 1977; Woodson et al. 1991) which aquires electrophysiological properties distinct from its descending (non-AcD) dendrite (Lischka et al. 2018; Weigel et al. 2022). Multiple axons can arise from different dendrites under conditions of pharmacological stabilization of microtubuli indicative of a substantial flexibility of neuronal polarity which is no longer dependent on centrosomes (Gomis-Rüth et al. 2008). The view is compatible with our observation of apical AcD as early as DIV 5. That major dendrites can give rise to axons or function as axon and as dendrite has also been demonstrated for the ‘dendrons’ of hypothalamic GnRH neurons (Herbison 2021). Also, dopaminergic neurons arise axons up to 100 µm distant to the soma (Moubarak et al. 2019). Whether or not AcDs could emerge much later e.g. during dendritic reorganisation or sprouting of additional dendrites cannot be answered with our approach. On average, about 20–25% neocortical pyramidal cells with basal AcD have been found in OTCs. Proportion resembles those seen in the non-primate neocortex in vivo (Wahle et al. 2022), in dissociated mouse hippocampal cultures (Han et al. 2025) and in ventral hippocampus and subiculum (Thome et al. 2014). Early deprivation and altering calcium reduced proportions up to DIV 20, and in these conditions basal dendritic length and branching was clearly impaired as reported (Gonda et al. 2020; Gasterstädt et al. 2020) and demonstrated by the present data. On the opposite, early enhancement of network activity increased the proportions at DIV 5 and DIV 10. It suggested an influence on the initial sprouting of basal dendrites because neither KA nor CCH mediated network activity could alter pyramidal cell basal dendritic complexity at DIV 10 and older (Jack et al. 2019; Wahle et al. 2026). Surprisingly, AcD proportions did not drastically overshoot. Rather, the controls appeared to catch up with time. Could this suggest that only a subset of pyramidal cells, for instance a certain transcriptional type (Yuste et al. 2020), can become an AcD neuron? Or, does an upper limit exist of how many basal AcD pyramidal cells a network can tolerate? Neurons with such privileged dendrites (Thome et al. 2014; Hodapp et al. 2022; Han et al. 2022) could lead to quite some noise in the circuit besides consuming metabolic energy. In hippocampus, the basal AcD configuration is essential for ripple oscillations. Visual cortex is not kown to express ripple oscillations, although in prefrontal cortex ripples can occur in relation to sleep phases and memory consolidation (Khodagholy et al. 2017; Shin and Jadhav 2024). Is it that the basal AcD phenotype is aquired in a regionally specific manner or even “on demand” to support specific network functions? The hippocampal 50% AcD neurons accounts for the CA1 region in mice at P30, while CA3 and subiculum display 20–30% AcD pyramidal neurons (Thome et al. 2014) which is closer to numbers seen in neocortex in vivo and in OTCs. The proportion of basal AcD pyramidal cells remained constant up to DIV 30 in OTCs. Also, similar proportions were seen in young and adult and aged mammals (Wahle et al. 2022). The decrease observed between DIV 7–21 in dissociated hippocampal cultures (Han et al. 2025) was not evident in cortical OTCs. No degeneration selectively of AcD neurons was observed. Possibly, over a life time, the set of AcD neurons is fixed. Alternatively, it may change such that today and in a week from now different pyramidal neurons present with AcDs e.g. by temporarily shifting the entire axon (Lehmann et al. 2023; Preprint) or for instance volume changes of the somata. From the earliest time point, 40–50% AcD interneurons were found in OTCs, similar to proportions seen in vivo (Höfflin et al. 2017; Wahle et al. 2022). The electrophysiology of AcD interneurons is hitherto unknown or has been rather sporadically assessed (Kaiser et al. 2001). Unexpectedly, none of the treatments influenced the proportion of BC and non-BC AcD cells, although GCaMP3 overexpression at DIV 14 slowed interneuronal dendrite growth (Gasterstädt et al. 2020) while early network activity and also GluK1 receptor activation enhanced interneuronal dendritic growth at DIV 10 (Jack et al. 2019; Wahle et al. 2026). Direct activation or inhibition of interneurons via GluN2D-containing NMDA receptors also failed to alter proportions of AcD interneurons although the manipulations affected the axonal branching in dependence on the axon origin (Köhler et al. 2025). We suggest that the AcD configuration of interneurons is entirely unrelated to initial dendritic growth. Further, the axon-is-first hypothesis does not account for interneurons. After immigrating from the lateral eminences via subplate/intermediate zone and marginal zone, interneurons sprout their local axons after arriving at their laminar destination, whereas pyramidal cells issue long-projecting axons already at the beginning of migration. Later however, interneuronal axons and AcDs reciprocally promote each other. The AcD becomes one of the longest dendrites, and specifically BC axons emerging from dendrites are ‘developmentally privileged’ in that they grow denser plexuses at DIV 10 and DIV 15 when compared to somatic BC axons (Gonda et al. 2023; Köhler et al. 2025; Wahle et al. 2026). It could suggest that somatic versus AcD occur by chance in interneurons an assumption which underscores how fundamentally different pyramidal cells and interneurons regulate their firing properties. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Acknowledgements We acknowledge support by the Deutsche Forschungsgemeinschaft Open Access Publication Funds of the Ruhr University Bochum. We thank Prof. Dr. Nicolai Bissantz, RUB, for statistical advice. Author contributions All authors contributed to the study conception and design. Culture preparation, transfection and staining was performed by D.B., B.C., A.R., I.K. and P.W. Data collection, analysis and visualization were performed by D.B. and P.W. The first draft of the manuscript was written by P.W. All authors commented on previous versions of the manuscript. All authors read and approved the final manuscript. This manuscript has been written to 100% by the human authors. Funding Open Access funding enabled and organized by Projekt DEAL. The authors declare that no funds, grants, or other support were received during the preparation of this manuscript. Data availability All data obtained during the study are presented in the manuscript. Declarations Competing interests The authors declare no competing interests. Ethics Statement All experiments were conducted in accordance with National and European (2010/63/EU) laws for the use of animals in research. Neonatal rats, born and housed in the local animal facility, were used for organotypic slice cultures. According to the German Animal Welfare Act (“Tierschutzgesetz der Bundesrepublik Deutschland, TierSchG”, § 4), the killing of vertebrates to remove tissue or organs for scientific purposes does not require official or institutional authorization. Therefore, no file number has been assigned to this project. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary Material 1 Data Availability Statement All data obtained during the study are presented in the manuscript.