Sensory axons induce epithelial lipid microdomain remodeling and determine the distribution of junctions in the epidermis

Abstract

Epithelial cell properties are determined by the polarized distribution of membrane lipids, the cytoskeleton, and adhesive junctions. Epithelia are often profusely innervated, but little work has addressed how neurites affect epithelial organization. We previously found that basal keratinocytes in the zebrafish epidermis enclose axons in ensheathment channels sealed by autotypic junctions. Here we characterized how axons remodel cell membranes, the cytoskeleton, and junctions in basal keratinocytes. At the apical surface of basal keratinocytes, axons organized lipid microdomains quantitatively enriched in reporters for PI(4,5)P2 and liquid-ordered (Lo) membranes. Lipid microdomains supported the formation of cadherin-enriched, F-actin protrusions, which wrapped around axons, likely initiating ensheathment. In the absence of axons, cadherin-enriched microdomains formed on basal cells but did not organize into contiguous domains. Instead, these isolated domains formed heterotypic junctions with periderm cells, a distinct epithelial cell type. Thus, axon endings dramatically remodel polarized epithelial components and regulate epidermal adhesion.

FIGURE 1:

Somatosensory axons induce epidermal layer–specific membrane remodeling. (A) (Top) Schematic of reporter expression in panels A–A′′′. The basal cell plasma membrane was labeled with a PIP2 reporter (Tg(tp63:GAL4VP16; UAS:EGFP-PLCδ-PH, green), and somatosensory neurons (magenta) were labeled with neural beta tubulin::DsRed or isl1:GAL4VP16; UAS:DsRed transgenes. (A–A′′′) Stills from a time-lapse movie. An extending neurite (cyan arrowhead) grew over the apical-facing surface of a basal cell. Shortly after the appearance of the axon, EGFP-PLCδ-PH–labeled membrane microdomains accumulated underneath the axon. Red arrow shows scattered microdomains that accumulated under an axon that had already grown over the basal cell. See Supplemental Movie 1. (B) (Top) Schematic of reporter expression in panels B–B′′′. Periderm cell plasma membranes (green) were mosaically labeled with a UAS:EGFP-PLCδ-PH transgene injected into Tg(krt5:GAL4FF); Tg(neural beta-tubulin:DsRed) embryos. Axons are labeled in magenta. (B–B′′′) Stills from a time-lapse movie. Axons had no apparent effect on the plasma membrane of overlying periderm cells. Cyan arrowhead marks a growth cone crawling underneath the labeled periderm cell in B–B′′ and the axon shaft after the growth cone has passed in B′′′. See Supplemental Movie 2.

FIGURE 2:

Basal cell membrane microdomains are enriched in a PIP2 reporter. (A) (Left) Schematic showing how relative reporter enrichment was quantified. Lipid microdomains are represented in gray, axon (not imaged) in blue. Intensity plots for EGFP-PLCδ-PH and mRubyCAAX were generated along a line drawn across microdomains (dotted black line). (Right) Intensity plot from a representative microdomain. To calculate relative enrichment, the maximum intensity within the microdomain was divided by the intensity of each membrane reporter just outside the microdomain. (B) Image of basal cell membrane coexpressing EGFP-PLCδ-PH and mRubyCAAX at 1 dpf. Microdomains were labeled by both reporters (arrowheads), but GFP was more enriched. mRubyCAAX also accumulated in numerous aggregates and/or endosomes in basal cells (asterisk). (C) Quantification of relative enrichment of the two membrane reporters in isolated microdomains (204 microdomains measured from 19 cells from seven embryos). Enrichment values from the same basal cell clones were averaged to obtain the average relative enrichment for each cell. EGFP-PLCδ-PH was significantly more enriched in microdomains than mRubyCAAX (p < 0.0001, paired t test).

FIGURE 3:

Axons reorganize epithelial PIP2 microdomains by lateral coalescence or extension from cell borders. (A–C′′′′) Stills from a Fast Airyscan movie of a 1 dpf Tg(tp63:GAL4VP16; UAS:EGFP-PLCδ-PH); Tg(neural beta tubulin:DsRed) embryo. Sensory axons are labeled in magenta and EGFP-PLCδ-PH in green. As growth cones (red and cyan arrowheads) from an axon first grow over a basal keratinocyte, the apical surface of the keratinocyte contains scattered PIP2 lipid microdomains (orange arrows). By 18–23 min, a subset of lipid microdomains have accumulated in the proximity of axons (orange brackets in A′′–C′′ and A′′′–C′′′). By 99.5 min, lipid microdomains have formed elongated axon-associated membrane domains (orange brackets in A′′′′–C′′′′). Images were acquired every 30 s. See Supplemental Movie 3. (D) Stills from Fast Airyscan movies of 1 dpf Tg(tp63:GAL4VP16; UAS:EGFP-PLCδ-PH); Tg(neural beta tubulin:DsRed) embryos showing different modes of lipid microdomain association with axons (arrowheads). (Left) Preexisting lipid microdomains coalesced underneath an axon segment. See Supplemental Movie 3. (Right) In other cases, an axon passing over a basal cell boundary was followed by PIP2-rich membrane from the cell borders. Images were acquired every 30 s. See Supplemental Movie 3. (E) Classification of microdomains into categories based on their behavior in 30 min movies (20 s imaging intervals). A total of 43.6% of microdomains were stable, persisting as isolated domains through the entire imaging interval; 25.7% of microdomains fused with other microdomains, while 9.6% formed by splitting from existing microdomains; 21.2% of microdomains were transient structures that disappeared during the course of the live imaging. Data from 15 cells (dots) from five embryos (color coded).

FIGURE 4:

Basal cells form F-actin protrusions around axons. (A) (Top) Stills from movies of 1 dpf Tg(tp63:GAL4VP16;UAS:EGFP-PLCδ-PH) injected with a 10xUAS:LifeAct-mRuby plasmid. The F-actin reporter signal (magenta) reorganized concomitantly with microdomain (green) remodeling. See Supplemental Movie 7. (Bottom) Linear intensity plots through the indicated cyan lines above showing the dynamic recruitment of F-actin to lipid microdomains. (B) Stills from a 1 dpf Tg(tp63:GAL4VP16;UAS:LifeAct-GFP);Tg(neural beta tubulin:DsRed) embryo taken by Fast Airyscan. During the initial outgrowth of sensory axons (magenta) into the skin, F-actin (green) recruited underneath axons dynamically wrapped axon segments (cyan arrowheads). See Supplemental Movie 8. Images were acquired every 30 s. (C) Two representative orthogonal stills from movies of the axon–basal keratinocyte interface from 1 dpf Tg(tp63:GAL4VP16;UAS:LifeAct-GFP);Tg(neural beta tubulin:DsRed) embryos imaged by Fast Airyscan (30 s intervals). Cyan lines in en face images indicate positions of orthogonal optical slices (bottom, “xz” panels). F-actin filaments (green) are recruited to the apical-facing cortex underlying axons (magenta) and surround the axon shaft (left, “1 min” still; right, “2.5” min still). Images were acquired every 30 s.

FIGURE 5:

Axon-associated domains progressively enrich reporters for Lo, but not Ld, membrane. (A) Double myristoylation and palmitoylation, or modification by glycosylphosphatidylinositiol anchors, localizes fluorescent protein reporters preferentially in Lo membrane. Gernanylgeranylated fluorescent reporters were used to label Ld membrane. (B) Schematic of membrane reporter localization (gray) with respect to axons (blue). Inset shows detail of axon-associated membrane domain (AAD). Relative enrichment of each membrane reporter was measured using the ratio of each reporter inside (“In”) to outside (“Out”) AADs. (C) Details of AADs in basal cells expressing Lo and Ld reporters bicistronically. Lo and Ld reporters were imaged between 1 and 5 dpf. At 1 dpf, relative enrichment was measured in both isolated microdomains (yellow arrowheads, “isolated” in D) and larger axon-associated domains (red asterisks, “elongated domains” in D) (D) Relative enrichment of AADs from basal cells expressing mp-mApple-T2A-mEGFP-gg. mp-mApple (Lo reporter) exhibited significantly greater relative enrichment than mEGFPgg (Ld reporter) at 1, 3, 4, and 5 dpf (**** p < 0.0001, Wilcoxon signed rank tests). Over time, the relative enrichment of mp-mApple increased significantly (*** p < 0.001, ANOVA with paired t test; **** p < 0.0001, ANOVA with Mann–Whitney U test). (E) Relative enrichment of AADs from basal cells expressing mp-mEGFP-T2A-mApple-gg or sfGFP-GPI-T2A-mApple-gg at 3, 4, and 5 dpf. Greater relative Lo enrichment was independent of fluorescent protein combination and can be recapitulated using an independent Lo reporter (sfGFP-GPI) (** p < 0.01, **** p < 0.0001, Wilcoxon signed rank tests). mp-mEGFP shows significantly greater enrichment at 5 dpf vs. 4 dpf (*** p < 0.001, ANOVA with Mann-Whitney U test) and vs. 3 dpf (**** p < 0.0001, ANOVA with Mann–Whitney U test).

FIGURE 6:

Lipid microdomains accumulate cadherins independently of axons. (A–B′′) To measure E-cad recruitment to lipid microdomain before axon association, Tg(tp63:GAL4VP16; UAS:EGFP-PLCδ-PH); cdh1-TdTomato embryos were injected with a control morpholino (Ctrl MO; see Materials and Methods) and imaged at 1 dpf. (A–A′′) Whole-cell view of a WT control basal cell with a PIP2 reporter (green) expressed specifically in basal cells and endogenous E-cad (magenta) expressed in both periderm and basal cells. The E-cad signal that does not overlap with the basal cell boundaries is from the overlying periderm layer. (B–B′′) Detail of basal cell marked by inset (cyan box in A) showing recruitment of E-cad (B′′) to PIP2 microdomains (B′) before their reorganization into ensheathment channels. (C–D′′) To measure E-cad recruitment to early microdomains in the absence of axons, Tg(tp63:GAL4VP16; UAS:EGFP-PLCδ-PH); cdh1-TdTomato embryos were injected with a neurogenin-1 morpholino (ngn1 MO) and imaged at 1 dpf. (C–C′′) Whole-cell view of ngn1 morphant basal cells. (D–D′′) Detail of PIP2 microdomains (D′) showing axon-independent recruitment of E-cad (D′′). Red arrows show an example of heterogeneous E-cad enrichment within a lipid microdomain. (E, F) To compare E-cad recruitment in microdomains in the presence (E) or absence (F) of axons, E-Cadherin-TdTomato and EGFP-PLCδ-PH intensities were measured across linear ROIs spanning microdomains (red dotted lines in B′ and B′′) and normalized to average fluorescence in the surrounding membrane. E-cad was weakly enriched in lipid microdomains before axon association (E), and this was unchanged by the absence of axons (F). (G–H′′) Representative images of 4 dpf Tg(tp63:GAL4VP16; UAS:EGFP-PLCδ-PH); cdh1-TdTomato larvae injected with control MO. E-cad recruitment to isolated microdomains (H–H′′, K) increased relative to 1 dpf (compare K with E). To measure E-cad recruitment in the prolonged absence of axons Tg(tp63:GAL4VP16; UAS:EGFP-PLCδ-PH); cdh1-TdTomato embryos were injected with a neurogenin-1 morpholino (ngn1 MO) and imaged at 4 dpf (I–J′′, L). PIP2 microdomains that formed in the absence of axons were still able to recruit and enrich E-cad to levels similar to those of microdomains that formed in WT basal cells (compare L vs. K). (Scale bars in A, C, G, I, 10 µm.)

FIGURE 7:

In the absence of axons, microdomains mature into stable periderm–basal contacts. (A) Tg(tp63:GAL4VP16;UAS:EGFP-PLCδ-PH); cdh1-TdTomato embryos injected with Ctrl MO or ngn1 MO and imaged at 3 dpf. In WT controls (A–A′′), E-cad in the basal layer localized to ensheathment channels (red arrows) and isolated microdomains (cyan arrowhead). In the absence of axons (B–B′′, C), basal cells formed an increased number of E-cad–containing isolated microdomains (p value computed using Welch’s t test, three embryos per condition). (D–F) One-minute-interval time-lapse movies of basal cell lipid microdomains were made using a conventional laser scanning confocal in Tg(tp63:GAL4VP16;UAS:EGFP-PLCδ-PH) embryos injected with control MO or ngn1 MO. See Supplemental Movie 9. In D, the overlap coefficient of each frame to the first frame of the movie is plotted over time to measure the dynamics of lipid microdomains unassociated with axons in four conditions. The data points (overlap coefficient = 1.0) of the first frame to the first frame is omitted in each condition. (E) Stills from representative movies are shown. Cyan arrowhead shows an example of a microdomain unassociated with ensheathment channels (red arrows). At 1 dpf, isolated lipid microdomains were highly dynamic structures, while at 4 dpf they were static structures, resulting in higher overlap coefficients. See Supplemental Movie 9. (F) Comparison of endpoint overlap coefficients for individual basal cells for each condition (p values computed using Mann–Whitney U tests. Ctrl MO 1 dpf, n = 35 basal cells from five embryos; ngn1 MO 1 dpf, n = 43 cells from four embryos; control MO 4 dpf, n = 45 cells from four embryos; ngn1 MO 4 dpf, n = 40 cells from four embryos).

FIGURE 8:

Axons inhibit the formation of periderm–basal cell junctions. (A) Periderm membrane (magenta) was specifically labeled using the krt5 promoter (Tg(krt5:mp-mCherry)), and basal membrane (green) was specifically labeled using the tp63:GAL4VP16 driver (Tg(tp63:GAL4VP16; UAS:EGFP-PLCδ-PH)) in larvae injected with ngn1 MO to block sensory neuron development. Images show xz view of the entire epidermis. Arrowheads show examples of overlapping microdomains at the periderm–basal boundary. (B–D′′) En face view of another periderm–basal boundary in a maximum-intensity projection. Boxes (C–D′′) show examples of overlapping lipid microdomains at the periderm–basal boundary. (E) Quantification of the proportion of basal cell microdomains that overlapped with periderm microdomains. Each dot (n = 56) represents an independent basal–periderm cell pair, measured in four larvae. In three out of four larvae (gray, pink, magenta dots), ∼50–100% of lipid microdomains in the two epithelial layers overlapped; in one larva (blue dots), <50% of lipid microdomains in the two skin layers overlapped, possibly due to differences in timing of junction maturation or developmental delay. An average of 24 microdomains were scored per cell pair. (F–G′) Representative TEM images of periderm–basal interfaces in 54 hpf WT (F) or ngn1 morphants (ngn1 MO, G). Junctions were identified as electron-dense plaques at the membrane (red bars in schematics on right). In WT epidermis, periderm–basal cell junctions were observed alongside wrapped axons (Ax in schematic). (H) Quantification of the density of periderm–basal junction plaques in WT and ngn1 morphant TEM skin sections. In the absence of axons, the density of ultrastructurally defined periderm–basal junctions increased (p = 0.0012, unpaired t test). Each data point represents a unique region of the epidermis from at least two larvae per condition.