Research ArticleNEUROSCIENCE

Social reprogramming in ants induces longevity-associated glia remodeling

See allHide authors and affiliations

Science Advances  19 Aug 2020:
Vol. 6, no. 34, eaba9869
DOI: 10.1126/sciadv.aba9869
  • Fig. 1 Single-cell transcriptomes from worker and gamergate brains.

    (A) Scheme of the experiment. Workers and gamergates were separated on the basis of behavior and ovary status. Brains were dissected and optic lobes removed. The central brain, including mushroom bodies (dark green), ellipsoid bodies (green), fan-shaped bodies (yellow), and antennal lobes (blue), plus the gnathal ganglion (purple) were dissociated into a single-cell suspension and processed for single-cell RNA-seq. (B) Annotated tSNE visualization of the clustering of 18,583 single-cell transcriptomes obtained by pooling all cells from six worker and five gamergate replicates. The number of cells in each cluster is indicated in parenthesis. IPC, insulin-producing cells. (C) Selected marker genes for the clusters annotated in (B). The y axis shows the collapsed pseudobulk expression in each cluster (as % of total cluster UMIs) for the indicated gene. Bars represent the means of 11 biological replicates + SEM. (D) Heatmap plotted over global tSNE showing normalized UMIs per cell for known neuronal markers (red) and glia markers (blue). (E) Heatmap for normalized expression levels (z score) for the indicated transcription factors (TFs) in collapsed single-cell clusters. Only transcription factors with a |log2(neurons/glia)| > 1 are shown, but the columns were clustered on all transcription factors. Astro A–C, astrocytes A–C.

  • Fig. 2 Distinctive features of Harpegnathos neurons and glia.

    (A) Annotated tSNE visualization for the reclustering of neurons from six worker and five gamergate replicates at day 30. (B) Heatmap plotted over neuronal tSNE showing normalized UMIs for known mushroom body markers from Drosophila (left) and A. mellifera (right). (C) Heatmap plotted over neuronal tSNE for the KC marker Pka-C1 from two Drosophila single-cell RNA-seq datasets, one from the central brain after removing optic lobes [left; (15)] and one from the whole brain, inclusive of the optic lobes [right; (14)]. (D) Relative abundance of KCs as determined by percentage of neurons in clusters that express Pka-C1 in Harpegnathos brains and in the two Drosophila single-cell RNA-seq datasets. Horizontal bars indicate means ± SEM. (E) Immunofluorescence for the neuronal marker synapsin and the KC marker Pka-C1 in Harpegnathos (left) and Drosophila (right) with 4′,6-diamidino-2-phenylindole (DAPI) as nuclear counterstain. Gray arrowhead, pedunculus; white arrowhead, calyx. (F) Western blot (WB) for Pka-C1 in the indicated amount of total protein extract from Drosophila (Dmel) or Harpegnathos (Hsal) brains. Tubulin was used as loading control. (G) Annotated tSNE visualization for the reclustering of glia from six worker and five gamergate replicates at day 30. (H) Clustered heatmap showing the pairwise Pearson correlation score between collapsed transcriptomes (pseudobulk analysis) of glia clusters, considering only variable genes that were used to define the clusters by Seurat. (I) Relative abundance of key glia subsets in Harpegnathos single-cell RNA-seq, Drosophila single-cell RNA-seq from the whole brain (14), and as defined by glial subset-specific expression of GAL4 (20). Bars indicate means + SEM.

  • Fig. 3 Cellular remodeling in the Harpegnathos brain after the caste transition.

    (A) Heatmap plotted over global tSNE showing the changes in cell type abundance in workers versus gamergate brains. The color scale indicates the effect size (mean/SD). The three clusters that are significantly (adjusted P < 0.1, Wald test) affected are indicated. (B) Volcano plot of log2(ratio) and −log10(adjusted P value from Wald test with Benjamini-Hochberg multiple test correction) for the relative frequency of each cluster in workers versus gamergates. Thresholds for false discovery rate (FDR) < 0.1 and FDR < 0.05 are shown. (C) Visualization and quantification of worker (left) and gamergate (right) contributions to the ensheathing glia cluster in the reclustered tSNE for glia only (see Fig. 2G). Worker and gamergate datasets were downsampled to include the same number of total cells for comparison. (D) Expression levels (% of cluster UMIs) across glia subsets for egr, Nep5L, and Slc5eg. Bars indicate means + SEM. (E) Immunofluorescence of a thick section of a Harpegnathos brain stained with antibodies against egr (red) and synapsin (green) and counterstained with DAPI (blue). The bottom panels show a magnification of the antennal lobes.

  • Fig. 4 Brain injury causes activation of ensheathing glia.

    (A) Scheme of the experiment. Needle-stabbed brains and age-matched control were analyzed by Drop-seq at days 1 and 3 after injury. (B) Annotated tSNE visualization of glia-only reclustering from pooled control and injury samples at days 1 and 3 (n = 3 control and injury for day 1; n = 3 control and 4 injury for day 3). The injury-specific cluster is indicated. (C) Relative frequency of cells in the injury-specific cluster as % of glia. Horizontal lines indicate means ± SEM. (D) Clustered heatmap showing the pairwise Pearson correlation score between collapsed transcriptomes of glia clusters considering only variable genes that were used to define the clusters by Seurat. The high transcriptome-wide correlation between the injury-specific cluster and resting ensheathing glia is indicated by a red square. (E) Expression levels (as % of cluster UMIs) for the activation marker Mmp1. Bars indicate means + SEM. (F) Head of a 30-day-old Harpegnathos worker subjected to antenna amputation. Photo credit: Lihong Sheng, University of Pennsylvania. (G) RT-qPCR for Mmp1 mRNA in the left (light gray) or right (dark gray) brain hemisphere after antennal ablation. Bars represent means + SEM. P value is from one-way analysis of variance (ANOVA) and Holm-Sidak test.

  • Fig. 5 Differential decline of ensheathing glia during aging in Harpegnathos castes.

    (A) Visualization and quantification of the contributions from workers of different ages to the ensheathing glia cluster in the Drop-seq tSNE. Cells from each time point were downsampled to the same total cell number for comparison. (B) Dynamics of the ensheathing glia cluster during aging. Each point represents a biological replicate. The line connects the means and error bars indicate ± SEM. The mean for the gamergate sample is shown as a thicker bar. ***P = 5 × 10−5 from a Wald test with multiple test correction. (C) Ratio of Mmp1 mRNA (RT-qPCR) in the right versus left brain hemisphere in ants subjected to ablation of the right antenna (gray) or a mock treatment as control (white). Activation of ensheathing glia as measured by up-regulation of Mmp1 was compared in young (30-day-old, left) and old (120- to 150-day-old, right) individuals. Bars represent means + SEM. The dashed line indicates a ratio of 1, i.e., no difference between right and left hemisphere. **P < 0.01 from one-way ANOVA and Holm-Sidak test.

  • Fig. 6 Age-associated decline of ensheathing glia in Drosophila.

    (A) Heatmap of UMI levels per cell plotted over tSNE for CG9657 in Drosophila brains obtained at the SCope website ( with single-cell RNA-seq data from (14). (B) Abundance of CG9657 mRNA as determined by RT-qPCR (normalized to Rpl32) from dissected brains of day 5 (d5; young) and day 70 (old) Drosophila females. Bars represent means ± SEM. P value is from a Student’s t test. (C) MA plot of RNA-seq data from brains of young (day 5, n = 3) and old (day 70, n = 4) Drosophila females. All 4436 genes identified as cell type markers in (14) are shown. Differentially expressed (FDR < 1%) genes are shown in black. Differentially expressed (FDR < 1%) ensheathing glia marker genes are in red. (D) Proportion of indicated gene classes that were significantly up-regulated (red) or down-regulated (blue) in old versus young brains according to RNA-seq. **P <0.01 from one-sided Fisher’s test. (E) Visualization of ensheathing glia in young (top) and old (bottom) Drosophila brains using the GMR10E12-GAL4 (25) driver line crossed to a UAS-GFP reporter. Brains were costained with a repo antibody (red) to visualize all glia cells. The central brain region used for quantifications is outlined in the merged images with dashed lines. (F) Quantification of ensheathing glia cells (GFP+) in the central brain plotted as percentage of total glia cells (repo+). Each point is a biological replicate. Bars represent the means ± SEM. P value is from a Student’s t test. (G and H) Same as (E) and (F) but for line GMR56F03 (20).

Supplementary Materials

  • Supplementary Materials

    Social reprogramming in ants induces longevity-associated glia remodeling

    Lihong Sheng, Emily J. Shields, Janko Gospocic, Karl M. Glastad, Puttachai Ratchasanmuang, Shelley L. Berger, Arjun Raj, Shawn Little, Roberto Bonasio

    Download Supplement

    The PDF file includes:

    • Figs. S1 to S8

    Other Supplementary Material for this manuscript includes the following:

    Files in this Data Supplement:

Stay Connected to Science Advances

Navigate This Article