Changes in transposable elements expression in male and female mice liver throughout aging

Abstract

Aging has traditionally been studied through the lens of protein-coding genes, with a strong bias toward data derived from male organisms. As a result, the role of non-coding elements and potential sex-specific differences remains largely unexplored. Transposable elements (TEs), mobile sequences capable of altering genome structure and regulating gene expression, have recently gained attention for their roles in development and aging. However, despite this growing interest, key aspects of TE expression dynamics are still poorly characterized, particularly in female tissues. To address this gap, we analyzed TE expression in RNA-Seq liver tissue from male (8, 26, 60, 78, and 104 weeks) and female (3, 24, 48, and 72 weeks) mice. Our results reveal distinct TE expression trends between sexes. While previous studies report increased TE expression with aging, we identified a subset of TEs with decreasing expression over time, differing between males and females. We also observed inverse expression trends between a few TEs and their nearby genes, supporting a potential regulatory relationship. We identified TEs with changing expression (CE TEs) through age associated with nearby genes showing strong expression correlations (|ρ|≥ 0.6). In males, correlated genes such as Txnrd2, Mthfd1, and Dkk3 are involved in redox regulation, one-carbon metabolism, and Wnt signaling, respectively, while in females, Thrb and Cd55 are linked to metabolic regulation and immune protection. These associations suggest that TE activity may be functionally coupled to transcriptional programs relevant to liver physiology and aging. These findings highlight the importance of examining TE expression in both sexes and suggest their potential regulatory roles in age-related liver physiology.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1007/s11357-025-02065-y.

Introduction

Aging is a broad field of research due to the various associated molecular changes and is characterized by progressive physiological deterioration, increased mortality, and a greater propensity for diseases such as heart disease, cancer, and neurodegenerative disorders [1]. One area of research focuses on the differences in the aging process between males and females. This field has been biased by the predominant use of males in experimental studies and clinical trials [2, 3]. Females are frequently excluded due to hormonal fluctuations, particularly during menstruation [4]. Despite this, females tend to have a lifespan that is, on average, 18.6% longer than that of males in mammal species. In humans, females present a 7.8% advantage in lifespan [4, 5]. Among the existing theories of aging, the possible biological explanations for sex differences are linked to chromosomes and hormonal differences [4, 6–8]. However, the main focus has been on the study of protein-coding genes, while a significant portion of the genome (~ 45% in mammals) consists of Transposable Elements (TEs). TEs are mobile DNA sequences capable of rearranging within the genome, a process that can contribute to genomic instability and, ultimately, cellular senescence [9–11].

TEs have gained importance in studies on aging and longevity, as they have been recognized as key regulators of gene expression and genomic instability, both of which are crucial factors during aging. The effect of TEs on ageing was first proposed by W. Murray in 1990, whose hypothesis suggests that the activity of TEs increases with age. By changing their location, TEs could inactivate necessary genes, ultimately leading to cell death [12–14]. Experiments in mouse (M.musculus), fruit fly (D. melanogaster), and nematodes (C. elegans) have shown that transposition and the resulting DNA damage accelerate the aging process [14–18]. Additionally, the frequency of TE transposition increases in different tissues as the organism ages [19–21]. Finally, suppressing TE activity has been linked to an extended lifespan [22–26].

Given the growing research and evidence suggesting that TEs play a role in aging and in the differences in longevity in the organisms, it is important to describe the expression patterns of TEs over time in males and females. For this study, we used publicly available RNA-Seq data from mouse liver, as it covers a wide range of age time points. With this data, we analyzed TE expression patterns over time. Our results will provide insights into the activity of TEs over time in mouse liver tissue between males and females.

Materials and methods

Data resources

The genome sequence assembly and annotation files of the mouse species, Mus musculus, (GRCm39/mm39) were downloaded from UCSC Genome Browser [27]. Paired-end RNA-Seq data from female and male mouse liver control samples were obtained from NCBI’s SRA Bioproject PRJNA562525 and National Genomics Datacenter’s (NGDC’s) Genome Sequence Archive (GSA) Bioproject PRJCA002140, respectively. The female mice dataset has age periods at 3, 24, 48, and 72 weeks with 3 biological replicates at each age [28] (SRA Bioproject PRJNA562525). The 12 libraries range between 130 and 186 million 101 bp paired-end reads. The male mice dataset includes time points at 8, 26, 60, 78, and 104 weeks with 5 biological replicates at each age [29] (GSA Bioproject PRJCA002140). The 25 male libraries range between 41 and 65 million 150 bp paired-end reads. The importance of these data sets lies in the coverage of a wide range of age time points. Among the most comprehensive datasets currently available to the public, they include samples from both sexes and cover a wide range of ages, providing valuable insights into the dynamics of mouse liver tissue across the lifespan.

Data set selection and read preprocessing

Quality control of the reads was performed using FASTQC (v0.11.9) [30] and visualized with MultiQC (v1.19) [31]. Based on the results obtained, Trimmomatic (v0.39) [32] was used to trim the first 10 nucleotides (HEAD:10), to remove reads with an average quality score of 28 or less (AVGQUAL:28), and to remove reads of a length shorter than 50nt (MINLEN:50). After the trimming process, between 2% and 3% of reads were filtered in both males and females, respectively.

Count matrix of TEs

RepEnrich2 (v2.31) [33] was used to determine which TEs were expressed in each RNA-Seq set at each age point in males and females. The mouse genome GRCm39/mm39 and its respective Repeatmasker TE annotation file [34] were used together to obtain the expression of TEs at each age point (files from UCSC Genome Browser) [27]. The count matrices for TEs were generated for all ages and for both sexes from RepEnrich2 results by a bash script.

To improve the reliability of detection, we retained TEs detected in at least 3 out of 5 biological replicates in males, and in at least 2 out of 3 biological replicates in females. These thresholds were chosen to include TEs consistently expressed in the majority of biological replicates while avoiding overly restrictive criteria that would require expression in all samples, which could exclude biologically relevant but variably expressed elements. The resulting count matrices were merged in R (v4.2.3) [35] to obtain a combined table containing all age points and both sexes.

Count matrix of genes

We used STAR (v2.7.10a) [36] to align the RNA-Seq reads from each dataset to the mouse mm39 genome. We used the GeneCounts function from STAR to generate the gene count matrix from RNA-Seq files. As done for the TEs, the genes were filtered based on having counts in males in 3 out of the 5 biological replicates. For females, counts in 2 out of the 3 biological replicates were considered. In both sexes, a minimum of 5 counts per sample was required.

per gene was considered. The resulting count matrices were merged in R (v4.2.3).

Prediction and integration of genomic origins of expressed TEs

The most probable genomic origin of each expressed TE was identified using TEcandidates (v2.0.5) [37], which reconstructs RNA-Seq reads de novo and aligns them to the reference mouse genome (GRCm39/mm39) annotated with RepeatMasker. The analysis was performed independently for each age group, using a minimum coverage of 30% of the annotated element and a minimum assembled candidate length of 100 nucleotides.

For each sample, TEcandidates reports a single best candidate locus per TE, minimizing ambiguity. However, when merging the results across all age groups by sex, some TEs appeared with more than one predicted genomic origin, reflecting variation in locus assignment between samples. To resolve these cases, the TE locus located closest to the nearest annotated gene was selected, favoring potential cis-regulatory relationships. When a TE was reported as being close to more than one gene, all possible TE–gene associations were retained for completeness.

The resulting TEcandidates coordinates were intersected with the mouse gene annotation using Bedtools (v2.27.1) [38], considering a ± 10 kb window upstream and downstream of each gene to capture potential regulatory proximity. TEs located within gene bodies (distance = 0 bp) were prioritized, as these are more likely to be co-expressed with their host genes through read-through transcription or intron retention [39]. In cases where TE was equidistant to 2 different genes, we assigned it to the gene that was more represented across age groups.

Finally, the curated TEcandidates coordinates were integrated with RepEnrich2-derived expression data by matching TE class, family, and subfamily names. This unified dataset was used to define subsets of CE TEs (TEs showing significant expression changes over time) and DE TEs (Differentially Expressed TEs between two age points) and to evaluate potential relationships between TE expressions and nearby genes.

Normalization of TE, gene counts and batch effect

We employed the edgeR (v3.40.2) [40–42] package in R v4.2.3 [43] to normalize both gene and TE count matrices using the TMM method, accounting for library sizes and differences in sequencing depth across samples. The data for both sexes were derived from SRA and GSA Bioprojects conducted by Jackson Laboratory, where mice were maintained and treated under the same standardized conditions, minimizing potential batch effects related to differences in laboratory practices or animal care. While the number of libraries differed between males and females, TMM normalization allows that comparisons between groups are valid by adjusting for these differences [44–46].

Numerous strategies have been developed to detect and correct batch effects in large-scale omics datasets [47–50]. However, in this study, each Bioproject corresponds exclusively to a single biological sex, resulting in complete collinearity between “sex” and “batch.” Applying batch correction under such conditions would remove the biological variability of interest. Principal Component Analysis (PCA) confirmed that the primary source of variation corresponds to the biological factor “sex,” consistent with this design structure. All statistical tests described below were performed separately for each sex to avoid cross-group dependencies.

Minimizing false positives through FDR adjustment

In this study, we applied a more stringent significance threshold (FDR ≤ 0.005) to minimize the likelihood of false positives. This value implies that, on average, fewer than 5 out of every 1000 elements identified as significant are expected to be false positives, compared to 5 out of 100 under the traditional threshold of FDR ≤ 0.05 [51–54]. Although this criterion reduces the total number of identified TEs, it allows the analysis to focus on a more reliable and statistically robust set. Given that several of these TEs are located near genes involved in hepatic function and aging, it suggests that they may have a potential regulatory role that warrants further investigation.

Time-course analysis of TEs and genes

We performed a time-course analysis using the edgeR package (v3.40.2) [40–42] to assess changes in the expression of TEs and genes across different age points. The analysis is based on Generalized Linear Models (GLMs), which model the count data and account for biological samples and technical replicates variability. All expression values were previously normalized using the TMM method. For each TE and gene, we fitted a GLM that incorporated age as a continuous variable to detect significant expression trends over time. Statistical significance was assessed using likelihood-ratio tests, and p-values were adjusted for multiple testing using the Benjamini–Hochberg false discovery rate (FDR) method as implemented in edgeR. The correction was applied independently for each sex and within each analysis type to ensure comparability while avoiding cross-analysis dependencies. Unless otherwise noted, an adjusted FDR ≤ 0.005 was considered significant [51–54]. These are called TEs with changing expression (CE TEs).

For visualization, fitted log₁₀(CPM) (logCPM, logarithm base 10 of Counts per Million) expression values were rescaled (z-score) across each feature using the pheatmap package (v1.0.12) in R, allowing comparison of relative expression patterns across ages. Heatmaps were generated separately for each sex using these rescaled fitted values.

Correlation analyses between CE TEs with genes and age

To investigate the relationship between CE TEs and their neighboring genes, as well as their association with age, we used Spearman’s rank correlation coefficient (ρ). This non-parametric approach was selected because it captures both the direction and strength of relationships without assuming linearity [55].

For each CE TE, we computed two independent correlations: TE–gene correlation: between the fitted expression values of each CE TE and its closest gene. Positive correlations (ρ ≥ 0.6) indicate concordant expression trends (both increasing or decreasing with age), whereas negative correlations (ρ ≤ −0.6) denote inverse relationships, suggesting potential regulatory interference; TE–age correlation: between the fitted expression values of each CE TE and the age vector. Positive correlations indicate sustained up-regulation with age, negative correlations indicate sustained down-regulation, and weak correlations (|ρ|≤ 0.6) were interpreted as transient or inconsistent expression patterns. These correlations were calculated independently for males and females using the cor() function in R with the method set to “spearman”.

This approach provided a quantitative description of the temporal expression dynamics of CE TEs across aging. Additionally, graphs were generated to visualize the correlation values between TE and gene pairs more effectively. Both scatter plots of gene vs. TE expression values and those showing the correlation values were created using the ggplot2 package (v3.5.1) [56] in R.

Differential expression analysis of TEs

Differential expression analysis (DEA) of transposable elements was performed using count data generated by RepEnrich2 [33]. As described in the normalization section, data were normalized using the TMM method implemented in the edgeR package (v3.40.2). To focus on aging-related rather than developmental differences, juvenile samples were excluded from the analysis. Thus, the earliest adult ages were used as reference points for each sex: 26 weeks for males and 24 weeks for females. Each subsequent age group was compared independently against its respective reference, as follows: Males: 26 vs. 60, 26 vs. 78, and 26 vs. 104 weeks;

Females: 24 vs. 48 and 24 vs. 72 weeks. Differential expression was evaluated using edgeR’s glmTreat() function, which implements the TREAT (Testing Relative to a Threshold) method [57]. This approach tests each contrast against a defined biological fold-change threshold, identifying TEs that display both statistical significance and meaningful biological effect sizes. For this analysis, a minimum absolute log₂ fold-change (|FC|≥ 1.2 or |log₂FC|≥ log₂(1.2) ) was used as the biological threshold, as suggested by the manual (https://rdrr.io/bioc/edgeR/man/glmTreat.html). Statistical testing was performed using the glmTreat() function in edgeR, which evaluates whether the true fold-change significantly exceeds this specified threshold rather than applying a strict cutoff to the observed estimate. Consequently, some significant elements may exhibit estimated log₂FC values slightly below ± log₂(1.2) in the plots, reflecting the hypothesis-testing framework of TREAT. p-values were adjusted for multiple testing using the Benjamini–Hochberg FDR method, and corrections were applied independently for each sex and for each pairwise comparison. Transposable elements meeting both the TREAT and FDR criteria were considered differentially expressed. The results were visualized using mean–difference plots (MD plots) generated with the ggplot2 package (v3.5.1) in R.

Results

TE expression patterns in male and female mouse liver

RepEnrich2 identified a total of 1,159 transposable elements (TEs) expressed in both males and females. After applying the filtering criteria based on biological replicates (detected in at least 3 of 5 male samples and 2 of 3 female samples), the retained TEs were 1,007 in males and 1,146 in females (Supplementary File 1 Fig. 1).

Fig. 1.

Principal Component Analysis (PCA) of mouse liver tissue TE expression in male and female mice across five age stages: The first two principal components with the largest variance are depicted, capturing 66.6% (PCA1) and 6.53% (PCA2) of the total variance. Two distinct groups are observed: the left cluster represents male data (green dots) and the right cluster corresponds to female data (red dots)

PCA revealed that the first two components accounted for 73.13% of the total variance in TE expression (PC1 = 66.6%, PC2 = 6.53%; Fig. 1). Apparently, samples clustered primarily by sex along PC1. However, this separation likely reflects both biological and experimental structure, as male and female data were obtained from independent Bioprojects, resulting in complete collinearity between the factors sex and batch. Consequently, batch correction could not be applied without also removing biological variation related to sex.

When analyzed separately, the first two principal components explained 35.3% of the total variance in males (PC1 = 24.6%, PC2 = 10.7%) and 69.8% in females (PC1 = 50.3%, PC2 = 19.5%) (Supplementary File 1 Fig. 2A-B). In males, TE expression showed an age-related gradient along PC2, with younger samples (8 and 26 weeks) clustering toward the top of the plot and older samples (60–104 weeks) toward the bottom. In contrast, in females, PC1 captured the major age-associated variation, separating young samples (3 weeks) from older ones (48–72 weeks) along the horizontal axis (Supplementary File 1 Fig. 2A-B). This indicates that while TE expression progressively changes with age in both sexes, the structure of these transcriptional trajectories differs, with aging effects being more linearly organized in females and partially dispersed across components in males.

Fig. 2.

Stacked bars chart of CE TEs near genes and their number per class in females and males. The Y-axis represents the number of the CE TEs over time by class. The X-axis represents the TE location, whether it is in intronic regions of gene or in its surroundings upstream or downstream (± 10kbp)

Genomic origin and proximity of expressed and CE TEs

The genomic origin of expressed and CE TEs was established by integrating the coordinates predicted by TEcandidates with expression data from RepEnrich2. Although TEcandidates reports a single best candidate locus per TE for each sample, merging results across all ages revealed cases where some TEs were assigned to different genomic loci depending on the dataset. These cases were resolved by selecting the locus closest to the nearest gene, prioritizing potential cis-regulatory associations. When a TE was predicted to be near more than one gene, all possible TE-gene relationships were retained. Using Bedtools, the predicted loci were intersected with annotated genes considering a ± 10 kb window around gene boundaries and in the same strand. This analysis revealed that most expressed TEs were located within genes or in their vicinity, suggesting potential transcriptional coupling. In total, 1,007 and 1,145 TEs were expressed in males and females, respectively. Among them, 996 (males) and 1,133 (females) could be matched to genomic origins, with 44.48% (males) and 23.48% (females) showing multiple potential loci, 48.59% (males) and 39.19% (females) having a unique origin, and 6.93% (males) and 37.33% (females) lacking confident origin (Supplementary File 2).

For CE TEs, 35 were identified in males and 64 in females (Supplementary File 1 Fig. 1). Of these, most lacked a confidently assigned origin (54.3% in males, 60.9% in females), while 17.1% of CE TEs in males and 35.9% in females mapped to a unique locus (Supplementary File 2). These results highlight the inherent ambiguity of repetitive elements and show that, even after stringent filtering, a considerable proportion of CE TEs cannot be confidently traced to a single locus. Among CE TEs located near genes, we identified 16 elements in males and 25 in females (Supplementary File 1 Fig. 1). Intronic positions predominated in both sexes (males: 9/16, 56.3%; females: 16/25, 64.0%), followed by downstream (3/16, 18.8% and 6/25, 24.0%) and upstream (4/16, 25.0% and 3/25, 12.0%) (Fig. 2). Class distributions differed by sex: in males, LTRs were most frequent (10/16, 62.5%), with fewer SINE (3/16, 18.8%), DNA (2/16, 12.5%), and LINE (1/16, 6.3%); in females, LTRs (11/25, 44.0%) were followed by DNA (8/25, 32.0%) and LINE (6/25, 24.0%), with no SINE detected. At the family level, ERVK was most represented overall (12 total; 4 in males, 8 in females), followed by L1 (7 total: 1 in male, 6 in females) and ERV1 (6 total: 3 per sex), with additional contributions from ERVL (2 in males only), ERVL-MaLR (1 in male), hAT-Tip100 (3 females), hAT-Ac (1 in female), TcMar-Tigger (3; 1 male, 2 females), TcMar-Tc1 (1 male), and SINE-related Alu/B2/B4 (1 each; males) (Fig. 3). Notably, all female LINE (L1) elements were intronic (6/6). Counts reflect occurrences; specifically, MamTip2 appears twice in females because its genomic origin could not be unambiguously resolved, and both matches were retained.

Fig. 3.

Horizontal bar chart of the number of CE TEs near genes by class/family, in males and females. The Y-axis represents the classes and families into which the CE TEs are categorized. The X-axis represents the total number of CE TEs

Time-course analysis of TEs and genes

The time-course analysis provided a global view of the temporal behavior of TEs and genes across ages. From this analysis, we identified 35 CE TEs in males and 64 in females (Supplementary File 2). This definition does not imply a progressive change across all ages; rather, it indicates that expression significantly varied at one or more time points, either transiently or in a sustained manner. To characterize expression dynamics, we extracted the fitted (adjusted) expression values for each CE TE and its associated gene at every age point and visualized them using row-scaled heatmaps (Fig. 4). In many TE-gene pairs, expression trajectories appeared concordant over time, while a smaller subset exhibited inverse trends (Supplementary File 2). To quantify these relationships, we computed Spearman’s rank correlations between fitted CE TE and gene expression, using |ρ|≥ 0.6 as an effect-size descriptor (not for significance testing). With this criterion, female pairs exhibited 12/25 positive and 1/25 negative associations, whereas male pairs showed 7/16 positive and 3/16 negative correlations (Supplementary File 3). Negative correlations (ρ < 0) indicate that TE expression increases while gene expression decreases, suggesting potential inverse regulatory interactions.

Fig. 4.

Heatmap of the expression of CE TEs and associated genes in males and females over time. Rows represent CE TEs (left) and genes (right) and columns represent the sampled ages in female (A) and male (B). The colors indicate rescaled expression values by scale parameter from pheatmap package in R. The blue represents lower expression levels and red represents higher expression levels

Finally, to assess time-dependent expression orientation, we calculated Spearman correlations between fitted expression values and age for each CE TE, classifying them as sustained upregulated (ρ ≥ 0.6), sustained downregulated (ρ ≤  − 0.6), or transient/weak (− 0.6 < ρ <  + 0.6). In females, 20 of 64 CE TEs were sustained up, 12 sustained down, and 32 weak or transient; in males, 22 of 35 were sustained up, 8 sustained down, and 5 weak or transient (Supplementary File 3). These temporal trends emerged without any directional pre-filtering and collectively support a potential regulatory interplay between CE TEs and nearby genes throughout aging.

Impact of juvenile ages on CE TEs detection and sustained expression patterns

To isolate aging-related signals, we repeated the time-course analysis excluding juvenile ages (females ≥ 24 weeks; males ≥ 26 weeks). Under the same statistical settings used in the full-spectrum analysis (edgeR GLMs, BH FDR ≤ 0.005), the number of CE TEs dropped sharply. In females, 3 CE TEs were detected, all with assignable genomic origin: IAPEz-int|LTR|ERVK, ORR1E|LTR|ERVL-MaLR, and RLTR17B_Mm|LTR|ERVK. In males, 2 CE TEs were detected: ETnERV3-int|LTR|ERVK and MamGypLTR3a|LTR|Gypsy (Supplementary File 4). These findings indicate that early developmental samples contribute disproportionately to CE TE detection when the entire lifespan is considered, whereas only a small subset of elements shows robust age-associated modulation within adulthood.

CE TEs and gene partners linked to aging biology and liver function

In males, CE TEs were preferentially found near genes involved in mitochondrial redox and epigenetic control (Txnrd2, Mthfd1, Dkk3, Rbms3, Snhg11), and near loci supporting hepatic homeostasis and proliferative balance (Gimap5, Tfpi2, Extl1, Spc25). These functions map onto aging-relevant axes, oxidative stress, Wnt/ECM remodeling, and cell-cycle control, that are central to liver maintenance over time (Supplementary File 2).

In females, CE TEs localized next to genes governing metabolic and nuclear-receptor programs essential for liver physiology (Thrb, Rxra, Cyp2c67, Cyp51, Ces1c, Slc9a9, Slc4a4, Ppm1g), as well as immune protection (Cd55) and chromatin/RNA regulation (Rnf20, Zc3h11a, Nab1). Additional partners implicated structural or secretory functions (Fmn1, Sil1, Atp6v0a4) and general signaling (Rap1gds1) (Supplementary File 2). Taken together, these positional associations highlight convergent pathways of metabolic/NR signaling, oxidative epigenetic control, and immune tone that are compatible with age-aligned transcriptional dynamics in liver.

Among these, a subset of gene–CE TE pairs displayed strong expression correlations (|ρ|≥ 0.6), highlighting potential regulatory coupling between TE activity and gene expression (Table 1). These pairs include Txnrd2, Mthfd1, and Dkk3 in males, key mediators of oxidative and epigenetic regulation. In females, Thrb, Cd55, and Rnf20, which participate in metabolic homeostasis, immune control, and epigenetic control of chromatin structure and genome stability via H2Bub1, respectively (Table 1).

Table 1.

Gene-CE TE pairs with |ρ|≥ 0.6 (Spearman correlation) in male and female mouse liver

Sex	Gene Coordinates	Gene	Ref	Brief Function	Relation to Aging	Relation to Liver Function	CE TE Name|Class|Family	TE Location	TE Position Relative to Gene	TE Distance to Gene (bp)
Male	chr6:3,962,595–3,988,919	Tfpi2	[75–77]	Regulating inflammatory responses and tumor cell proliferation	Induce apoptosis and inhibit angiogenesis, which may contribute significantly to tumor growth inhibition	Regulates extracellular matrix degradation in the liver; overexpression is linked to cirrhosis and hepatocellular carcinoma	LTR16A1|LTR|ERVL	chr6:3,991,916–3,992,343	Upstream	2998
Male	chr7:111,715,224–111,758,264	Dkk3	[78–82]	Modulator of the Wnt signaling pathway, involved in development and cell differentiation	Wnt pathway alterations contribute to tissue degeneration and aging	Regulates liver regeneration and hepatocyte proliferation; dysregulation is linked to fibrosis and cancer	X33a_DNA|DNA|TcMar-Tc1	chr7:111,716,369–111,716,595	Intronic	0
Male	chr12:76,302,072–76366577	Mthfd1	[83–87]	Trifunctional enzyme in folate metabolism and the one-carbon cycle, essential for nucleotide synthesis	Impaired folate metabolism accelerates DNA damage and aging	Folate metabolism is crucial for liver function; deficiencies lead to fatty liver disease and impaired methylation	L1MEi|LINE|L1	chr12:76,365,643–76,365,875	Intronic	0
Male	chr16:18,245,134–18,297,823	Txnrd2	[88, 89]	Key antioxidant enzyme in peroxide reduction and protection against oxidative stress	Reduced activity increases oxidative damage, a key driver of aging	Maintains redox balance in hepatocytes; depletion is linked to liver fibrosis and oxidative stress-related liver diseases	MER31-int|LTR|ERV1	chr16:18,240,149–18240563	Upstream	4572
Female	chr1:130,366,764–130390481	Cd55	[90–93]	Regulator of the complement system, protecting cells from immune attack	Decreased expression with aging affects immune response and inflammation	Protects hepatocytes from immune-mediated damage and inflammation	ETnERV2-int|LTR|ERVK	chr1:130,361,471–130363867	Downstream	2898
Female	chr4:49,632,006–49656887	Rnf20	[94–96]	E3 ubiquitin ligase that catalyzes histone H2B monoubiquitination (H2Bub1), regulating chromatin structure, transcriptional elongation, and DNA damage repair	By maintaining H2Bub1 and supporting accurate double-strand break repair and epigenetic stability, RNF20 limits genomic instability and pro-oncogenic transcriptional programs	Reduced RNF20/H2Bub1 levels promote hepatic stellate cell activation, extracellular matrix accumulation, and fibrosis, whereas sustained RNF20 activity supports liver homeostasis and protects against chronic liver injury	X6b_DNA|DNA|TcMar-Tigger	chr4:49,663,004–49663219	Downstream	6118
Female	chr14:4,431,611–4,809,435	Thrb	[97–99]	Thyroid hormone receptor involved in metabolism regulation	Thyroid hormone signaling is deregulated with age, affecting metabolism	Regulates liver lipid metabolism, glucose homeostasis, and bile acid synthesis	L1MdV_I|LINE|L1	chr14:4,671,401–4677481	Intronic	0

Differentially expressed TEs (DE TEs)

Differential expression analyses anchored to early adulthood (26 weeks in males, 24 weeks in females) were performed using glmTreat (|log₂FC|≥ log₂(1.2); ≈0.26) with FDR correction. In males between 26 and 60 weeks, 2 TEs: MamGypLTR3a (LTR/Gypsy) and ETnERV3-int (LTR/ERVK) were upregulated at 60 weeks relative to the 26-week baseline. The 26 versus 78 weeks comparison revealed the largest number of differentially expressed TEs, comprising 4 upregulated and 5 downregulated elements. The upregulated set included ETnERV3-int (LTR/ERVK), RLTR26_Mus (LTR/ERVK), MamGypLTR3a (LTR/Gypsy), and L1MA8 (LINE/L1). The downregulated group comprised L3 (LINE/CR1), MLT1A0 (LTR/ERVL-MaLR), RLTR19 (LTR/ERVK), ERVB7_1-LTR_MM (LTR/ERVK), and RLTR6C_Mm (LTR/ERV1) (Fig. 5). At the latest time point, 26 vs. 104 weeks, ETnERV3-int (LTR/ERVK) remained upregulated, suggesting sustained activation of this locus across the aging trajectory.

Fig. 5.

Differential expression of transposable elements during adulthood in male and female. Mean difference (MD) plots showing the differential expression of transposable elements (TEs) across adult age contrasts obtained with edgeR glmTreat (|log₂FC|≥ log₂(1.2) = 0.26; FDR-adjusted). (A–C) Comparisons for males using 26 weeks as the reference age (26 vs. 60, 26 vs 78, and 26 vs. 104 weeks). (D–E) Comparisons for females using 24 weeks as the reference (24 vs. 48 and 24 vs. 72 weeks). Each dot represents a TE; red denotes significantly up-regulated elements, respectively, while black points indicate non-significant cases. Dashed horizontal lines indicate log₂FC = 0, and dotted lines mark the biological threshold (± 0.26).

In females, no differential expression was detected at 48 weeks relative to 24 weeks, but one TE, ORR1E (LTR/ERVL-MaLR), was upregulated in the 24 vs. 72 weeks comparison (Fig. 5). Collectively, these adult-anchored comparisons indicate that TE expression changes during adulthood are limited in number but detectable in males, with most DE elements belonging to LTR retrotransposons (ERVK, ERVL, Gypsy) and LINE (L1) families. The recurrence of ETnERV3-int and MamGypLTR3a across multiple contrasts points to their consistent upregulation during aging.

Discussion

In this study, we analyzed the expression of TEs that could be related to aging. Within these results, we observed changes in the expression of TEs over time and differences in TE expression between males and females. The results suggest that TEs could play a regulatory role in genes, and the differences found are markedly influenced by sex.

TE expression patterns across sexes

One of the most evident patterns in our data is that samples cluster primarily by sex rather than by age, as shown in the PCA analysis (Fig. 1; Supplementary File 1 Fig. 2). This pattern likely reflects both biological and experimental factors, since data for males and females were obtained from independent Bioprojects, making the factors sex and batch completely confounded. Consequently, direct comparisons between sexes are not appropriate, and all downstream analyses were performed independently for males and females.

Within each sex, however, we observed temporal changes in TE expression, suggesting that transposable elements may participate in regulatory processes that vary with age. Previous studies have reported that TEs can influence gene regulation during sexual development [58], and that sex hormones such as estrogen and testosterone can interact with TEs, modulating their transcriptional activity [59]. These mechanisms could underlie part of the sex-specific TE expression patterns observed across studies. Moreover, epigenetic differences between males and females may also contribute to the observed expression dynamics [60, 61, ].

Although the present study cannot separate biological and technical contributions to the overall sex-dependent clustering, the consistent trends detected within each sex provide valuable insight into how TE activity may be modulated over time. Future studies should include balanced experimental designs across both sexes to disentangle biological and batch-related variation and to deepen our understanding of how TEs contribute to gene regulation and aging.

TE expression and their potential regulatory influence on nearby genes

The change in TE expression could be related to alterations in chromatin conformation over time. One of the mechanisms that regulates TE expression is the compact (heterochromatin) and relaxed (euchromatin) states of chromatin. When chromatin is compact, it regulates the expression of genomic elements and prevents them from being expressed because the transcription machinery cannot bind. As the individual ages, it is possible that some regions of chromatin shift from heterochromatin to euchromatin, allowing the elements in the genome to be expressed [62]. It is known that epigenetic changes, such as DNA demethylation or histone acetylation, can facilitate this chromatin transition, which may contribute to the increased expression of certain TEs with aging [63–65].

In our study, we not only observed the expected increase in TE expression, as described in previous investigations [14, 19, 26, 66–68], but we also identified a subset of TEs whose expression decreases over time, in both males and females (Fig. 4, Supplementary File 1 Fig. 3). Although this does not happen in the same TEs and genes or to the same extent between both sexes, the same phenomenon is observed in both males and females. Similar to the TEs that increase their expression over time, those TEs that decrease in expression are mostly found in intronic regions and a few in intergenic regions at a close distance to genes (10kbp upstream or downstream), potentially influencing their expression. Studies have shown that TEs can act as regulatory elements, such as enhancers or even alternative promoters, where both TEs and genes are expected to increase their expression [12]. Additionally, we observed CE TEs that exhibit a negative correlation with gene expression, meaning that if CE TE expression increases, gene expression decreases, and vice versa. This was observed in both males and females (Supplementary File 1 Fig. 4).

These contrasting patterns may reflect two non-exclusive biological scenarios. On one hand, certain TEs may act as regulatory elements (enhancers, promoters, or silencers) whose activity changes with age, modulating nearby gene expression either positively or negatively. On the other hand, the subset of TEs showing decreasing expression over time may represent residual transcription from developmentally active elements that progressively lose expression as the tissue matures and reinforces its epigenetic identity. During embryonic and early postnatal stages, several TE families such as ERVL–MaLR and ERVK exhibit broad transcriptional activity associated with chromatin remodeling and cellular plasticity, which gradually diminishes as epigenetic repression becomes established [69].

Taken together, these observations suggest that TEs may play dual roles during lifespan, some gaining regulatory activity through epigenetic relaxation with age while others diminish in expression as remnants of developmental transcription that are gradually silenced during tissue maintenance.

CE TEs and associated genes related to aging and liver function

In this study, we identified CE TEs positioned near genes with well-established roles in aging-related processes and liver homeostasis (Table 1). The strong correlations (|ρ|≥ 0.6) between CE TEs and these genes support the idea that age-associated TE activity may be functionally coupled to pathways central to hepatic integrity, rather than randomly distributed across the genome.

In males, CE TEs localize next to Tfpi2, Dkk3, Mthfd1, and Txnrd2, aligning with axes of extracellular matrix remodeling and tissue protection (Tfpi2), Wnt-mediated regulation of regeneration and fibrosis (Dkk3), one-carbon/folate metabolism and DNA methylation (Mthfd1), and mitochondrial redox defense (Txnrd2). These genes integrate inflammatory control, epigenetic balance, and oxidative stress responses which are core hallmarks of liver aging and susceptibility to chronic injury.

In females, CE TEs are associated with Cd55, Rnf20, and Thrb, linking TE dynamics to complement regulation and immune protection (Cd55), epigenetic control of chromatin structure and genome stability via H2Bub1 (Rnf20), and thyroid hormone–dependent metabolic regulation (Thrb). Together, these genes converge on immune surveillance, fibrosis modulation, and systemic metabolic tuning, all critical for preserving liver function across age.

Although the causal mechanisms remain unresolved, the non-random proximity and coordinated behavior of CE TEs with these aging- and liver-relevant genes highlight CE TEs as candidate regulatory elements embedded within key hepatoprotective and pro-degenerative pathways. Future work should test whether these CE TEs directly modulate gene expression or instead mirror broader transcriptional and epigenetic shifts accompanying liver aging.

Potential trans-regulatory effects of CE TEs.

Although our analyses focused on cis-proximal associations, several CE TEs showed no coordinated expression changes with their nearest genes or did not exhibit a strong correlation with the adjacent gene, suggesting the possibility of trans-regulatory effects. These include, in males, LTR67B|LTR|ERVL–C130026I21Rik and MMERVk10C-int|LTR|ERVK–A530064D06Rik, and in females, MARE10|DNA/TcMAR-Tigger–Atp6v0a4, MMVL30-int|LTR|ERV–Or5m3b, RLTR41|LTR|ERV1–Gm21839, and X9a_DNA|DNA–Gm9959. The lack of parallel or strongly correlated expression patterns in these pairs implies that CE TE activity might not be limited to cis-regulatory interactions but may also involve trans-regulatory mechanisms, such as TE-derived RNAs acting on distal genes, long-range enhancer activity, or chromatin remodeling across genomic domains [70, 71]. Future studies integrating long-range chromatin mapping (Hi-C/HiChIP), nascent transcription assays (GRO-seq/PRO-seq), or targeted perturbations (CRISPRi/CRISPR-deletion) will be essential to disentangle local from distal TE regulatory impacts during aging.

LTR elements as the predominant class among age-associated CE TEs

Although LTR elements are not the most abundant TE class in the mouse genome, they appear to be the most transcriptionally active during aging. This can be explained by their high CpG methylation sensitivity: during aging, loss of DNA methylation at these loci facilitates their derepression and transcriptional reactivation [72, 73]. In addition, LTRs contain endogenous promoter and enhancer sequences, remnants of their retroviral origin, that can autonomously initiate transcription or act synergistically with nearby genes [74]. In contrast, LINE and SINE elements depend on the retrotransposition machinery or co-activation by L1, making them less responsive to moderate epigenetic drift. Moreover, LTRs tend to be evolutionarily younger and structurally more intact, preserving long terminal repeats and internal open reading frames that support their transcriptional competence [14]. Together, these features could explain why LTRs are overrepresented among CE TEs and likely drive the majority of age-related TE expression changes.

Implications for aging and future research

The results regarding the expression of TEs near genes in liver tissue suggest that TEs may regulate gene expression during aging, particularly in processes related to immune regulation, metabolic homeostasis, and chromatin remodeling. The inverse correlations observed between TEs and neighboring genes suggest a potential regulatory role, which justifies further investigation. Future studies should prioritize the functional characterization of these elements through epigenetic analyses (e.g., ATAC-seq, ChIP-seq) to determine whether TE activation is accompanied by changes in chromatin accessibility or histone modifications. The deletion or silencing of specific TEs using CRISPR could help evaluate their direct impact on gene regulation. In addition, single-cell transcriptomics could be used to explore whether TE–gene interactions occur in specific hepatocyte subpopulations or at certain developmental stages.

Finally, our results reinforce the relevance of analyzing TE expression in both male and female models. The sex-specific differences observed in TE expression dynamics and associated biological processes suggest that TE regulation may contribute to sex-based variation in liver aging and in the susceptibility to age-related diseases.

Study limitations

Our study relies on publicly available RNA-Seq datasets, which constrain the analyses to the sequencing methods and data characteristics of each source. Despite differences in the number of biological replicates and the age range covered in each dataset, these resources represent some of the best currently available for investigating transposable elements (TEs), as both used paired-end RNA-Seq with read lengths exceeding 100 bp. This improves alignment accuracy and reduces multi-mapping issues, which are common when mapping highly repetitive sequences.

However, an important limitation is that the male and female datasets were obtained from independent Bioprojects, leading to complete collinearity between the biological factor sex and the batch effect. Consequently, batch correction could not be applied without also removing sex-related biological variability. In addition, the age windows between sexes do not fully overlap (males: 8–104 weeks; females: 3–72 weeks), which further limits the interpretation of direct sex-specific effects. Therefore, all analyses were conducted independently for males and females, and observed differences between sexes should be interpreted cautiously.

We recommend that future aging studies include both sexes sampled under unified experimental conditions and across matched time points to enable proper disentanglement of sex- and age-related effects. Another limitation concerns the limited benchmarking and consistency among software tools for TE expression analysis. Although various algorithms have been developed, the associated estimation error remains relatively high [39]. Finally, our study lacks direct experimental validation of the regulatory impact of TEs on nearby genes. Nevertheless, the TEs and genes identified here represent potential candidates for future functional studies aimed at corroborating their possible regulatory relationships and their role in the biology of aging.

Conclusions

In this study, we developed an analytical strategy to investigate the behavior of TEs and identify those showing significant temporal changes in mouse liver. By conducting parallel analyses for male and female datasets, we characterized TEs’ dynamics within each sex independently, due to the complete collinearity between sex and batch. This approach allowed us to explore temporal trends and potential regulatory roles of TEs while avoiding confounding between biological and technical variation.

Our results revealed sets of TEs that increase or decrease their expression over time, suggesting that some elements may act as regulatory modulators of nearby or host genes. Furthermore, a subset of TEs showing both differential and time-dependent expression was found near genes with well-characterized biological functions, supporting their potential contribution to gene regulation during aging.

These findings underscore the importance of including both sexes in future experimental designs under conditions where batch and biological factors can be independently controlled to better understand the contribution of TEs to gene regulation and to the molecular mechanisms underlying the aging process.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contribution

BHR proposed the research, made the main analyses, wrote the first version and subsequent improvements of the manuscript; PM contributed to the discussion and understanding of the results from the biological and experimental point of view; GR contributed to the statistics, suggesting data analyses, providing computational resources, writing part of the manuscript, and approving the final version.

Funding

Chilean Government Science Funding Agency ANID’s Fondecyt Grant #1231357 to GR; ANID’s Doctoral fellowship #21201090 to BHR.

Data availability

The data analyzed in this study are available for download from NCBI’s Sequence Read Archive (SRA), under Bioproject PRJNA562525 and from National Genomics Data Center’s (NGDC’s) Genome Sequence Archive (GSA), under Bioproject PRJCA002140 (see Data Resources section in Materials and Methods) in the following links: https://www.ncbi.nlm.nih.gov/bioproject/PRJNA562525

https://ngdc.cncb.ac.cn/bioproject/browse/PRJCA002140

Declarations

Ethics approval and consent to participate

Not applicable.

Conflict of interest

The authors declare no conflict of interest.