Sequence Diversity and Encoded Enzymatic Differences of Monocistronic L1 ORF2 mRNA Variants in the Aged Normal and Alzheimer's Disease Brain

Overview of the experimental and bioinformatic pipeline for PacBio long-read sequencing and analysis of the L1 neural transcriptome. Schematic of the experimental workflow and bioinformatic pipeline for analysis of the poly(A⁺) L1 neural transcriptome. Sample preparation: PFC and MTG of ND and AD postmortem human brain (ages 46–94 years; n = 31). RNA extracted and gDNA contamination removed through DNase treatment and columns. cDNA synthesized with oligoDT primers to isolate poly(A⁺) RNA transcripts. Library enriched for L1 through custom Twist Bioscience pulldown probes. L1 enriched libraries barcoded to allow sample identification and then sequenced on PacBio Sequel II. Data pre-processing: ∼23 million long reads obtained from the PacBio Sequel II were used to generate high-quality consensus reads. Barcoded adapters were removed, and proper read orientation was determined to generate full-length nonconcatemeric reads (∼16 million high-quality reads; 72.3%). L1 detection: L1 detected via dual methods—Censor (identification, annotation, and masking via the human reference library of repeats) and minimap2 alignment to the consensus L1 sequence to ID “L1+ sequences” (Brouha et al., 2003). L1 quantification: Censor annotation and masking utilized to identify L1 families within each read and the presence of flanking regions mapping back to the reference genome (“read-through” transcripts). L1-containing transcripts examined for prevalence of promoter and regulatory regions. Open reading frames identified via getORF and then aligned to consensus ORF1 and ORF2 sequences from UniProt [Q9UN81; O00370] via BLASTp, allowing L1+ transcripts to be assigned into subcategories: ORF1 alone, ORF1 + ORF2, ORF2 alone, ORF1 + partial ORF2 (ORF2P), and ORF2P alone. Variants called if supported by ≥3 reads.

PacBio long-read sequencing of the L1 neural transcriptome reveals interindividual and sample variation. A, Histogram of the percentage of L1 reads of different lengths (bp-basepairs). Bin 500 bp. Median ± interquartile range (IQR). The red square indicates L1 reads ≥6 kb (∼0.77%). B, The percentage of fragmented reads for transcripts of each length based on assessment of captured non-L1 and non-HERVk containing reads as incomplete-splice matches via SQANTI3. Median, IQR, min, and max. C, The percentage of L1 reads assigned to the youngest subfamily of L1, L1PA. Mean + SD. D, The percentage of L1 reads assigned a L1 family annotation based on Censor. Mean + SD. Reads were frequently annotated with multiple subfamilies (average of 1.86 annotations/read)—percentages of read assignments total to >100%. E, The number of variants identified and number of samples containing at least one variant per category. F, Number of reads (per 10,000 L1 reads) identified as containing ORF1 alone, partial ORF2 (ORF2P) alone, ORF1 + ORF2P, ORF2 alone, or ORF1 + ORF2. Dots represent individual samples. Numbers indicative of mean. Error bars indicate SD. G, The percentage of variants identified in only one sample, indicating high interindividual variability in variant expression. H + I. The number of variants for categories (H) ORF1 + ORF2, ORF2 alone, ORF1 + ORF2P, ORF2P alone, and (I) ORF1 alone per individual sample for AD versus ND samples and PFC versus MTG samples. J–N, Pile-ups of reads for ORF1 alone (J), ORF2P alone (K), ORF1 + ORF2P (L), ORF2 alone (M), and ORF1 + ORF2 (N) as aligned to the L1 consensus sequence. O, The percentage of L1 coding categories (ORF1 alone, ORF2P alone, ORF1 + ORF2P, ORF2 alone, ORF1 + ORF2) reads assigned each L1 family annotation based on Censor. The average number of annotations per read for each coding category noted above each column. P, The percentage of reads within each L1 coding category with flanking regions that aligned to regions of hg38 with annotated genes, indicative of intragenic, “read-through” L1 transcripts.

RT activity in the human brain correlates with L1 ORF2 in neurons. A, Schematic of the experimental workflow and FPERT assay. B. RT activity (as pU of control recombinant HIV RT) of postmortem human brain samples from ND and AD PFC (filled circles) and MTG (open circles). Median ± interquartile range (IQR), min and max, nonsignificant (ns): Mann–Whitney test. C, Relative RT activity of the microdissected gray versus white matter compared with whole brain lysates from three brains. Mean ± SD. **p < 0.01; ***p < 0.001; ****p < 0.0001. Multiple unpaired t test. D, Scatterplot of RT activity (pU) in postmortem human brain samples (n = 31) relative to ORF2 H-score in MAP2 + nuclei. *p = 0.0176; y = 12.22 × −357.3; simple linear regression. E, Scatterplot of RT activity (pU) in postmortem human brain samples (n = 31) relative to ORF1 H-score in MAP2+ nuclei. ns. Simple linear regression. F, Schematic of RT expression vector transfection and activity assessment via FPERT assay. G, FPERT activity assessment of RT protein lysates (ORF2, HERVk-pol, and H-TERT) compared with transfection controls (EGFP—background cell lysate RT activity) and RT activity-positive controls (10pU HIV RT). Delta Cq: negative control Cq (no RT, no signal after 50 cycles)—sample Cq; n = 3, mean, SD; one-way ANOVA with Tukey–Kramer; ****p < 0.0001.

Nucleotide and amino acid variation across L1 ORF2 variants for overexpression-based function assays. Nucleotide and amino acid variants across the 12 functionally assayed ORF2 variants as compared with the ORF2 region in the L1 consensus sequence (Brouha et al., 2003). EN (yellow), RT (blue), and C-terminus (red) domains are highlighted. Percentage identity, percentage coverage, and the number of mismatched nucleotides and amino acids as compared with the L1 consensus sequence.