Widespread and heterogenous expression of ORF1p protein in the mouse brain

(A) Schematic representation of the unbiased cell detection pipeline on large scale and confocal images. Immunofluorescent images on sagittal mouse brain slices were acquired on a digital pathology slide-scanner or on a confocal microscope (DNA stain: Hoechst, neuronal marker; NeuN, protein of interest: ORF1p). Pyramidal images aquired with the slide scanner were then aligned with the hierarchical anatomical annotation of the Allen Brain Atlas using ABBA. Once the regions defined, a deep-learning based detection of cell nuclei (Hoechst staining, Stardist) and cell cytoplasm (NeuN staining, Cellpose) was performed on each sub-region of the atlas. Objects were filtered according to the background intensity measured in each sub-region for each channel (NeuN and ORF1p). The identity and intensity measures were analyzed at the regional and whole brain level. In parallel, confocal images (multiple z-stacks) of two selected regions (frontal cortex and ventral midbrain) were also acquired and identity and intensity were quantified using Cellpose and Stardist.

(B) Widespread and heterogenous expression of the LINE-1 encoded protein ORF1p in the mouse brain. Representative image of ORF1p immunostaining (orange) of a sagittal section of the brain of a young (three months-old) mouse acquired on a slide scanner. Scale bar = 1mm. (1-10) Representative images of immunostainings showing ORF1p expression (orange) in 10 different regions of the mouse brain acquired on a confocal microscope. Nuclei are represented in blue (Hoechst), scale bar = 50µm. (1) Isocortex, (2) Hippocampus, (3) Striatum dorsal, (4) Thalamus, (5) Midbrain motor, (6) Pallidum, (7) Hypothalamus, (8) Substantia nigra pars compacta, (9) Hindbrain, (10) Cerebellum. ORF1p expression profile in the mouse brain. The entire mouse brain with the exception of the olfactory bulb and the cerebellum were analyzed according to the pipeline on large-scale images described in (A). Bar plot showing the total number of ORF1p+ cells per mm² in the mouse brain. Data is represented as mean ±SEM, n=4 mice (top). Bar plot indicating the proportion of ORF1p+ cells compared to all cells detected. Data is represented as mean ±SEM, n=4 mice, 202001 total cells analyzed (middle). Scatter plot showing the mean intensity of ORF1p per ORF1p+ cell. Data is represented as mean ±SD, n=4 mice, 40999 ORF1p+ cells analyzed (bottom).

(D-F) ORF1p expression profile (density, proportion and expression) in defined anatomical regions of the mouse brain. Nine anatomical regions as defined by the Allen Brain Atlas and mapped onto sagittal brain slices (four three-month-old Swiss/ OF1) with ABBA were analyzed using the pipeline on large scale images described in (A). (D) ORF1p+ cell density in 9 different regions. Bar plot showing the number of ORF1p+ cells per mm². Data is represented as mean ±SEM; *p<0.05; **p<0.01; adjusted p-value, one-way ANOVA followed by a Benjamin-Hochberg test (E) Proportion of ORF1p positive cells in 9 different regions. Bar plot showing the proportion of ORF1p+ cells among all cells detected per region. Data is represented as mean ±SEM. (F) Mean ORF1p expression per cell in 9 different regions. Dot plot showing the mean intensity of ORF1p signal per ORF1p+ cell in 9 different regions. Data is represented as mean ±SD. The number of analyzed cells per region is indicated in the figure. *p<0.05; **p<0.01; ***p<0.001; ****p<0.0001; adjusted p-value, nested one-way ANOVA followed by Sidak’ multiple comparison test.

(G) ORF1p expression in the frontal cortex and ventral midbrain. Confocal images with multiple z-stacks were analyzed using Cellpose and Stardist. Dot plot representing the mean expression of ORF1p per ORF1p+ cells. Individual (four three-month-old Swiss/ OF1) mice are represented each by a different color, the scattered line represents the median. ****p<0.0001, nested one-way ANOVA. Total cells analyzed = 4645.

(H-I) ORF1p expression in the frontal cortex and the ventral midbrain. (H) Western blots showing ORF1p (top) and actin expression (bottom) in four individual mice per region which were quantified in (I) using actin as a reference control. The signal intensity is plotted as the fold change of ORF1p expression in the ventral midbrain to ORF1p expression in the frontal cortex. *p<0.05; two-sided, unpaired student’s-test.

(J) ORF1p expression in three regions of the human brain. Western blot showing ORF1p expression in the cingulate gyrus (CG), frontal cortex (FC) and cerebellum (CB) of post-mortem tissues from a healthy individual. ORF1p (Top), Actin (bottom).

ORF1p is predominantly expressed in neurons in the mouse brain

(A) ORF1p expression is absent in the white matter (corpus callosum) and predominantly expressed in neurons. Proportion of ORF1p+/NeuN+, ORF1p+/NeuN-, ORF1p-/NeuN+ and ORF1p-/NeuN-cells in the white matter (corpus callosum) and the grey matter (left) and in nine different regions (right) analyzed by the cell detection pipeline on large scale images presented in Figure1A. Exact values can be found in Suppl_Table1.

(B) Representative confocal microscopy images showing ORF1p (red) and NeuN expression (green) in two different regions of the mouse brain. The bottom images show the merge of the two stainings, an overlap of both markers is represented in orange. z-projections; scalebar = 25µm.

(C) Proportion of neurons expressing ORF1p in the frontal cortex and ventral midbrain quantified on confocal images. ns: non-significant; chi-square test on the cell number of the different cell-types analyzed; n=4 mice, data is represented as mean ± SEM.

(D) Proportion of ORF1p+ cells identified as NeuN+ or NeuN- in two different regions, analyzed by confocal microscopy on multiple z-stacks. ns: non-significant; chi-square test, n=4 mice.

(E) ORF1p does not colocalize with glial or microglial cell markers. Representative confocal microscopy images showing ORF1p staining (red) and three different glial cell (GFAP, Sox9, S100β) or microglial (Iba1) markers (white). Note that Iba1 antibody (rabbit) was used with the ORF1p 09 antibody (guinea pig, in house) z-projections, scalebar = 50µm.

(F-G) Separation of neuronal and non–neuronal - cells by FACS confirms predominant neuronal expression of ORF1p. (F) Neuronal (NeuN+) and non-neuronal (NeuN-) cells isolated by fluorescent activated cell sorting (FACS). Dot plots showing autofluorescence versus an appropriate control antibody (IgG rabbit 647; left) and an antibody against NeuN (name of the AB 657, right). The P4 window represents isolated NeuN+ cells (pink) and the P5 fraction NeuN-cells (orange) containing the same number of cells as sorted in P4 for comparison, others NeuN- are represented in blue. (G) Western blot. ORF1p expression (top), in NeuN- and NeuN+ FACS-sorted cells stemming from Figure F.

(H) Representative confocal microscopy images showing ORF1p (red), NeuN (green) and Hoechst (blue) in the cingulate gyrus of the human brain. z-projection; scalebar = 25µm (left). Example of individuals neurons expressing ORF1p or not are shown on the right panel. z-projection; scalebar = 5 µm (right).

(I) Proportion of ORF1p+ cells identified as NeuN+ or NeuN- in the human cingulate gyrus, analyzed by confocal microscopy on multiple z-stacks.

(J) Proportion of neurons expressing ORF1p in the human cingulate gyrus, analyzed by confocal microscopy on multiple z-stacks.

ORF1p expression is increased throughout the whole mouse brain in the context of aging

(A) Proportion of ORF1p+/NeuN+, ORF1p+/NeuN+, ORF1p+/NeuN-, ORF1p-/NeuN+ and ORF1p-/NeuN-cell-type in the whole brain (left) and the different analyzed regions (right) of young (three-month aged, n=4) and aged (16-month-old, n=4) mice using the cell detection pipeline on large scale images presented in Figure 1A, data is represented as mean ± SEM. Exact values can be found in Suppl_Table1.

(B) Proportion of ORF1p+/NeuN+, ORF1p+/NeuN+, ORF1p+/NeuN-, ORF1p-/NeuN+ and ORF1p-/NeuN-cell-type in two different regions of young and aged mice, analyzed on multiple z-stack confocal images. ns: non-significant; *p<0.05 calculated using two-way ANOVA with sidak’s multiple comparisons test on the cell number of the different cell-types analyzed; data is represented as mean ± SEM.

(C) ORF1p mean expression per ORF1p+ cell in the brain analyzed on large-scale images. Dot plot showing the ORF1p mean expression per ORF1p+ cell in young (n=4) and aged (n=4) mice in the whole brain (except cerebellum and olfactory bulb). 74985 total cells were analyzed; * p<0.05, two-way ANOVA with sidak’s multiple comparisons; data is represented as mean ± SEM.

(D) Frequency distribution of ORF1p mean intensity in ORF1p+ cells. ***p<0.001, Kolmogorov-Smirnov test.

(E) Frequency distribution of Hoechst mean intensity in the nuclei of OrF1p+ cells. ns: non-significant, Kolmogorov-Smirnov test.

(F) Mean ORF1p expression per ORF1p+ cell in nine different anatomical regions. Dot plot showing the ORF1p mean expression per ORF1p positive cell (n=74985). Adjusted p-value are represented, two-tailed nested t-test followed by a Benjamin, Krieger and Yukutieli test; n=4 young and n=4 aged mice per region, data is represented as mean ± SEM.

(G) Color-coded representation of fold-changes of ORF1p expression with aging. Represented is the fold-change in percent (aged vs young) of the “mean of the mean” ORF1p expression per ORF1p+ cell quantified mapped onto the nine different regions analyzed as shown in (F).

(H) ORF1p expression is increased in the ventral midbrain of aged mice. Dot plot representing ORF1p expression in two different regions of young and aged mice analyzed on confocal images with multiple z-stacks; total cells analyzed = 8381 ns: non-significant *p<0.05, two-tailed one-way ANOVA; dashed lines represent the medians.

(I) Representative confocal microscopy acquisition showing increased ORF1p expression (red) in the ventral midbrain region of aged mice (one z plan is shown). Cell nuclei are shown in blue (Hoechst staining). Scalebar = 50µm.

Young LINE-1 elements are increased in aged human dopaminergic neurons

TE transcript expression in RNA-seq data of laser-captured micro-dissected post-mortem human dopaminergic neurons of brain-healthy individuals was analyzed using RepeatMasker (multimappers) or the L1Base (unique reads).

(A) Volcano plot of differential analysis of LINE-1 expression using DESeq2 comparing young (≤65y, n=6) or aged (>65y, n=35) human dopaminergic neurons at the “name” level of RepeatMasker. Young LINE-1 elements, including the two families L1HS and L1PA2 that have coding copies, are highlighted in red.

(B) Scatter plots of normalized read counts (“name” level) of the young L1HS and L1PA2 families as well as the human endogenous virus family HERVK-int, another TE family with coding potential comparing young (≤65y, n=6) or aged (>65y, n=35) human dopaminergic neurons. Mann-Whitney test, p<0.05.

(C) Correlation of the expression of LINE-1 elements with known regulators in human dopaminergic neurons. Spearman correlation of evolutionary close (L1HS, L1PA2-17) and distant LINE-1 (L1PB and L1MA5) as well as HERV elements with coding potential (HERV-Kint, HERV-Fc1, HERV-Fc2 and HERV-H-int) with known regulators of their expression. HERV-W and TREX1 did not pass the normalized read count threshold of >3 reads in >6 individuals.

Dysregulation of locus-specific full-length LINE-1 elements in aged human dopaminergic neurons

(A) Volcano plot of differential expression analysis of TE expression using DEseq2 comparing young (≤65y, n=6) and aged (>65y, n=35) human dopaminergic neurons at the locus-level of specific full-length LINE-1 elements (140 of 146 “UID’s” as annotated in L1Base; threshold >3 reads in at least 6 individuals).

(B) Pairwise comparison of the expression of 140 out of 146 full-length LINE-1 elements comparing young (≤65y, n=6) and aged (>65y, n=35) human dopaminergic neurons. Wilcoxon matched signed rank test, p<0.0001left panel).

(C) The sum of read counts of all UIDs per individual were plotted comparing young (≤65y, n=6) and aged (>65y, n=35) human dopaminergic neurons.

(D-E) Dysregulated locus-specific full-length LINE-1 elements (UID-68 and UID-129) are plotted as scatter plots comparing young (≤65y, n=6) and aged (>65y, n=35) human dopaminergic neurons. (D) UID-68 is located adjacent to the genes CLEC5A and OR9A4 (left). Spearman correlation analysis of the expression of UID-68 and CLEC5A (middle) or OR9A4 (right) in young (≤65y, n=6, black dots) or aged (>65y, n=35, red squares) human dopaminergic neurons. (E) UID-129 is intergenic.

Endogenous ORF1p interactors in the mouse brain

Immunoprecipitation (IP) of endogenous ORF1p from the mouse brain. WB against ORF1p showing ORF1p enrichment after IP but no signal in the IgG control. Five independent samples were then prepared for proteomic analysis by mass spectrometry (LC-MS/MS).

(B) GO slim enrichment analysis of proteins selected as endogenous ORF1p protein partners in the mouse brain after quantitative LC-MS/MS. ORF1p-immunoprecipitated proteins were categorized into GO slim terms. The nine GO slim term with the highest fold-change are plotted. Fold enrichment is depicted on the upper axis and displayed as bars, the FDR value appears on the lower axis and is represented by the black points. BP: Biological Process, CC: Cellular Component, MF: Molecular Function.

(C) Venn diagram showing common interactors (purple) between interactors of endogenous ORF1p in the mouse brain identified in this study (red) and known (published) interactors of ORF1p (blue). Statistical significance of the overlap between the two groups of proteins was tested by an overrepresentation test (http://nemates.org/MA/progs/overlap_stats.html).

(D) ORF1p associates with the SWI/SNF complex (red), RNA pol II complex (orange) and interactors belonging to GO terms related to neuronal cell body & neuron projection (green). Known interactors previously published 60,65–70 are indicated with a purple ring. STRING network of physical interactions where nodes represent proteins partners identified in (A) and edges thickness represents the strength of shared physical complexes. Only proteins sharing physical interactions were represented.

(A) Selective recognition of ORF1p antibody. IHC showing ORF1p positives cells in sagittal mouse brain slice (left) and abolition of the signal when blocking the antibody with purified ORF1p (right).

(B) Representative acquisition showing ORF1p obtained with the widely used, commercially available ORF1p ab antibody (abcam ab216324) used in this study (red) and with an in-house ORF1p gp antibody (guinea pig, green) in mouse brain. Scalebar = 20µm (top) and 100µm (bottom).

(C) Quantification of double positives (gp+/ab+) cells using ORF1p ab antibody (abcam ab216324) and in-house ORF1p gp antibody (guinea pig) versus single-positive cells (gp+/ab- and gp-/ab+) in mouse frontal cortex (left) and ventral midbrain (right).

(D) ORF1p is expressed in six different brain regions in the mouse.

Brain regions were micro-dissected from a three-month old mouse brain. Western blot showing ORF1p expression in 6 brain regions. ORF1p (Top), Actin (bottom).

(A) Proportion of neurons (NeuN+) expressing ORF1p in different regions of the mouse brain as quantified using the cell detection pipeline on large scale images.

(B) ORF1p is predominantly expressed in neurons. Proportion of ORF1p+/NeuN+, ORF1p+/NeuN+, ORF1p+/NeuN-, ORF1p-/NeuN+ and ORF1p-/NeuN-cells, ****p<0.0001; calculated using chi-square test on the cell number of the four different cell-types analyzed by confocal microscopy on multiple z-stacks.

(C) ORF1p cell identity. Proportion of ORF1p+ cells identified as NeuN+ (black) or NeuN- (grey), in the whole brain (left) and in 9 different regions analyzed (right) using the cell detection pipeline on large scale images presented in Figure 1A; data is represented as mean ± SEM, n=4 mice.

(D) Proportion of neurons in the frontal cortex and ventral midbrain quantified using confocal approach. *p<0.05, chi-square test on the cell number of the different cell-types analyzed; n=4 mice, data is represented as mean ± SEM.

(E) Representative slide-scanner acquisition of a human cingulate gyrus section showing NeuN positives cells (green) mostly located in the grey matter (right) compared to the white matter from a brain-healthy individual; scale bar = 400µm. Zoom into the grey matter region showing ORF1p is presented in Figure 2H.

(A) Proportion of ORF1p+ cells being neuronal in the ventral midbrain comparing young and aged mice as quantified using confocal approach. Kolmogorov-Smirnov test; data is represented as mean ± SEM.

(B) Proportion of neurons expressing ORF1p in the ventral midbrain comparing young and aged mice as quantified using confocal approach. Kolmogorov-Smirnov test; data is represented as mean ± SEM.

(C) Proportion of neurons in the ventral midbrain comparing young and aged mice as quantified using confocal approach. Kolmogorov-Smirnov test; data is represented as mean ± SEM.

(A-B) Comparison of the expression of dopaminergic markers tyrosine hydroxylase (TH, A) and LMX1B (B) between young (≤65y, n=6)) and aged (>65y, n=35) human dopaminergic neurons. Mann Whitney test.

(C, E) Volcano plot of differential expression analysis of TE expression using DEseq2 comparing young (≤65y, n=6) and aged (>65y, n=35) human dopaminergic neurons at the “class” (C) and “family” (E) level of RepeatMasker.

(D) Scatter plot comparing the expression of LINE at the “class” level between young (≤65y, n=6)) and aged (>65y, n=35) human dopaminergic neurons. Mann Whitney test.

(F) Scatter plot comparing the expression of LINE and Alu at the “family” level between young (≤65y, n=6)) and aged (>65y, n=35) human dopaminergic neurons. Mann Whitney test.

(G) Scatter plot comparing the expression of HERVH-int, HERV-Fc1 and two non-coding, non-autonomous but active TEs in the human genome, AluYa5 and SVA-F at the “name” level between young (≤65y, n=6)) and aged (>65y, n=35) human dopaminergic neurons. Mann Whitney test.

(A-C) Correlation analyses of L1HS expression with Engrailed 1 (EN1, A, Spearman r=-0.43, p=0.002), CBX5/HP1 (B, Spearman r=-0.35, p=0.01) and XRCC6 expression (C, Spearman r= −0.394, p=0.005). Normalized read counts are plotted. Black dots correspond to young individuals (≤65y), red dots correspond to aged individuals (>65y).

(D-G) Scatter plots comparing the expression of EN1, CBX5/HP1, XRCC5 and XRCC6 between young (≤65y, n=6)) and aged (>65y, n=35) human dopaminergic neurons. Student’s t-test (EN1) or Mann Whitney test (HP1, XRCC5/6).

(A) Correlation of mappability (UMAP hit counts over UID, see methods) and UID expression (normalized read counts). Spearman correlation.

(B-D) In silico analysis of annotated full-length LINE-1 elements as in L1Basev2 (human reference genome hg38). (B) Percentage of L1HS and L1PA2 elements among the 146 full-length elements (UID1-146). (C) Percentage of full-length LINE-1 elements located inside or outside a gene. (D) Presence (blue, with gene symbol) or absence (white) of a “hosting” gene among the 146 annotated full-length LINE-1 in the human reference genome

(E) Mean expression of all 146 full-length LINE-1 elements in dopaminergic neurons of all individuals ≤65y.

(A-C) Dysregulated locus-specific full-length LINE-1 elements are plotted as scatter plots comparing young (≤65y, n=6) and aged (>65y, n=35) human dopaminergic neurons. (A) UID-37 is located in an intron of the gene HPSE2 (left). Spearman correlation analysis of the expression of UID-37 and HPSE2 in young (≤65y, n=6, black dots) and aged (>65y, n=35, red squares) human dopaminergic neurons. (B) UID-127 is located within the 6th intron of the non-coding RNA LINC00598 (left). Spearman correlation analysis of the expression of UID-127 and LINC00598 in young (≤65y, n=6, black dots) and aged (>65y, n=35, red squares) human dopaminergic neurons. (C) UID-137 is intergenic.