Transcriptome-wide piRNA profiling in human brains for aging genetic factors

Abstract

Objective:

Piwi-interacting RNAs (piRNAs) represent a molecular feature shared by all nonaging biological systems, including the germline and somatic cancer stem cells, which display an indefinite renewal capacity and lifespan-stable genomic integrity and are potentially immortal. Here, we tested the hypothesis that piRNA is a critical genetic determinant of aging in humans.

Methods:

Expression of transcriptome-wide piRNAs (n=24k) was profiled in the human prefrontal cortex of 12 subjects (84.9±9.5, range 68–100, years of age) using microarray technology. We examined the correlation between these piRNAs’ expression levels and age, adjusting for covariates including disease status.

Results:

A total of 9,453 piRNAs were detected in brain. Including seven intergenic and three intronic piRNAs, ten piRNAs were significantly associated with age after correction for multiple testing (|r|=0.9; 1.9×10⁻⁵≤p≤9.9×10⁻⁵).

Conclusion:

We conclude that piRNAs might play a potential role in determining the years of survival of humans. The underlying mechanisms might involve the suppression of transposable elements (TEs) and expression regulation of aging-associated genes.

Introduction

Piwi-interacting RNAs (piRNAs) represent the largest class of non-coding RNAs. Initially discovered and most abundant in germline cells, piRNAs have also been identified in somatic tissues, including the brain, heart, small intestine, kidney, liver, lung, skeletal muscle and pancreas ^1,2. Recently, we reported 9,453 piRNAs in human brains and 8,759 in human stomachs, with 103 and 50 piRNAs associated with Alzheimer’s disease (AD) and gastric cancer, respectively ³. On the basis of functional versatility of piRNAs, we postulate that piRNAs may be related to more phenotypes, including years of survival, in humans.

As the organism ages, transposable elements (TEs) in the genome of somatic cells multiply and become increasingly active and mobile ^4,5, most obviously in neurons ^6–8. TEs are capable of moving from one genomic locus to another, thereby causing insertional mutations ^9,10. Both the TE-derived mutation rates ^7,8,11–14 and the mortality rates ¹⁵ increase exponentially over the adult lifespan, in contrast to the constant rates in association with TE-independent mutations and chemical or physical mutagens ¹⁴, suggesting a decisive mutagenic role of TEs during aging. A recent animal study suggested that aging was related to many TE-related genes ¹⁶.

TE-derived insertions consist of normal, chemically unaltered nucleotides, which are difficult to be recognized and eliminated by effective DNA repair mechanisms ¹⁴, thus constituting the primary source of genomic instability, a molecular hallmark associated with aging ¹⁷. Genomic instability triggered by unrepaired mutations limits cellular capacity in proliferation and self-renewal, and contributes substantially to the aging process ^14,17. As a result, TE-generated molecular damages may overwhelm the capacity of cellular maintenance and DNA repair systems as the organism ages, leading to degenerative pathologies and eventually organism death ^14,18. There are two major classes of TEs, as distinguished by the mechanism of transposition ⁹ – RNA transposons (herein called retrotransposons or mobile elements) and DNA transposons. The retrotransposons include long terminal repeat elements (LTR), long interspersed nuclear element (LINE), and short interspersed nuclear elements (SINE). TEs exert pathogenetic effects via controlling the transcription of neighboring genes.

piRNA represents a shared feature of all nonaging biological systems, including the germline, somatic cancer stem cells, and certain ‘lower’ eukaryotic organisms that all display an indefinite renewal capacity and a lifespan-stable genome integrity and are potentially immortal ¹⁴. The sequences of piRNAs primarily complement the TEs. Thus, the primary function of piRNAs is to suppress the activity of TEs, which protects against TE-mediated mutagenesis, maintains genomic integrity, self-renewal ability and proliferation capacity and, as a result, promotes years of survival ¹⁴. A recent animal study suggested that up-regulation of piRNA expression could probably explain the long life expectancy ¹⁶. This function of piRNAs is much more efficient than other small interfering RNAs (siRNAs) ^19,20. The genomic locus transcribing the piRNA cluster collects single copies of all TE families, serving as a ‘genomic memory’ or immunity-like system that has evolved for genome maintenance by suppressing deleterious TE activity ¹⁴. Other non-piRNA mechanisms in preserving genomic stability have not been identified. Additionally, piRNA may harbor a heretofore unexplored, TE-independent mechanism to ensure genomic integrity by regulating gene transcription via modulation of chromatin organization ^14,21. piRNA may also interact with Piwi protein and then bind to the non-TE regulatory regions of aging-associated genes, resulting in deadenylation and inactivation of these genes and eliminating a lifespan-limiting pathology ¹⁴.

In the present study, we examined the relationship between years of survival and piRNAs across the transcriptome in humans and explored the potential mechanisms underlying the piRNA-longevity relationship.

Materials and Methods

Subjects and piRNA expression profiling:

The prefrontal cortex tissues dissected out from the brains of 12 subjects were studied. All brain tissues were obtained from the Human Brain Bank in Chinese Academy of Medical Sciences & Peking Union Medical College, which collects brains from donors through a whole-body donation program ^22–24. The age of death of the donors ranged from 68 to 100 years (mean 84.9±9.5 yrs) (Table S1). The postmortem intervals (from the time of death to brain tissue emersion fixation or freezing) were less than 13 hours (mean 6.0±3.0 hrs) (Table S1), which were shorter than some international brain banks, e.g., Douglas Montreal Brain Bank ²⁵. All donors or their guardians had given informed consent before death of donors for using the donated body tissue for medical research.

The 12 subjects included 6 with Alzheimer’s disease (AD) (2 males and 4 females) and 6 controls free of neuropsychiatric diseases (2 males and 4 females). Diagnosis of the subjects was confirmed both clinically and pathologically, based on National Institute on Aging (NIA) Alzheimer’s Association guidelines for the neuropathologic assessment of Alzheimer’s disease ²⁶. Subjects were classified according to the ABC score, a score of AD neuropathologic changes that incorporates (A) histopathologic assessments of amyloid β deposits, (B) staging of neurofibrillary tangles, and (C) scoring of neuritic plaques. The ABC scores reflect four levels of AD neuropathologic change: Not, Low, Intermediate or High. Subjects with Intermediate or High ABC scores were classified as cases with AD, and subjects with low or no ABC scores were classified as controls (Table S1). Control subjects showed no evidence of any neurodegenerative disorder during their lifetimes and were matched to cases in age (±5 years), sex, ethnicity and postmortem interval (Table S1). The detailed demographics and neuropathological criteria of the sample have been published elsewhere ^3,22–24. The present study was approved by the Institutional Review Board of the Institute of Basic Medical Sciences, Chinese Academy of Medical Sciences (Approval Number: 009–2014, 031–2017).

Briefly, transcriptome-wide piRNAs (n=24k) were profiled and validated for the prefrontal cortexes of 12 human subjects using microarray technology and qPCR. The piRNAs incorporated in the established Arraystar Human piRNA Microarray (Arraystar, Inc., Rockville, MD, USA) have been verified by interacting with Piwi proteins.

piRNA-longevity correlation analysis:

We performed Pearson correlation on piRNA expression and years of survival. A total of 9,453 (40% of the 24k) piRNAs were detected in prefrontal cortex ³ and located in 500 piRNA clusters (http://www.smallrnagroup.uni-mainz.de/piRNAclusterDB.html). piRNAs within the same cluster were highly correlated. Therefore, the significance level (α) for piRNA-longevity associations was set at 10⁻⁴ (=0.05/500).

The pathological changes, including synaptic degeneration and tangle formation, may affect piRNA expression in the brain; that is, if a piRNA is associated with AD, it might be impacted by the AD-related neuropathological changes. Therefore, the associations between AD and longevity-associated piRNAs were also analyzed. If this association was significant (p<0.05), disease status would be set as a covariate in the piRNA-longevity association analysis using linear regression analysis, to control for the confounding effects of potential neuropathological changes.

mRNA and protein expression of risk protein-coding genes in the prefrontal cortex:

The primary function of piRNAs is to suppress TEs that regulates the transcription of protein-coding genes. Considering the complex DNA tertiary structures, the possibility of the TEs to trans-regulate distant genes cannot be completely excluded. However, without prior information regarding the tertiary structures, the TEs are most likely to cis-regulate nearby genes. Thus, we only examined the proximate genes here. piRNAs affect the phenotypes primarily via TE suppression. The target TEs that are potentially regulated by the risk piRNAs were searched from UCSC Genome Browser (http://genome.ucsc.edu; genome assembly, hg19; track, RepeatMasker). In addition to TE suppression, piRNAs might also regulate the proximate protein-coding genes via non-TE pathways ^14,21. The expression of these genes in brain or other tissues is critical for piRNAs to exert their effects on years of survival (Figure S1). We therefore carefully examined the expression levels of their mRNAs or proteins in four independent cohorts, including one brain expression database (BRAINEAC) and three whole-body expression databases (BioGPS, GTEx and ProteinDB; detailed elsewhere ²⁷) that included a total of 190 tissues including the prefrontal cortex (Table S2) and other tissues (data not shown). The subjects in these four cohorts were free of any neuropsychiatric or neurodegenerative disorder, as confirmed both clinically and pathologically ^28–33. If the genes were not most abundant in brains, the tissues where they were most abundant are listed in Table 1.

Table 1.

Significant associations between piRNAs and longevity

piRNA	chromosome position	gene [organ of most abundant expression]	mean intensity (normalized) (log2 transformed)	Correlation* with longevity		Association with AD
				r	p	p
DQ581919	chr3:12936932–12936960	close to IQSEC1 [brain (mRNA and protein)]	15.0	0.9	9.9×10−5	0.18
DQ594437	chr6:33860547–33860575	close to GRM4 [brain (mRNA)]	10.4	0.9	7.0×10−5	0.27
DQ593237	chr15:51584181–51584210	intron 1 of CYP19A1 [Placenta (mRNA and protein) or ovary (mRNA)]	4.0	−0.9	4.7×10−5	0.68
DQ575255	chr15:51300659–51300685	intron 1 of CYP19A1 [Placenta (mRNA and protein) or ovary (mRNA)]	10.4	0.9	7.4×10−5	0.21
DQ597687	chr15:102293704–102293735 (telomere)	close to TM2D3 [brain (mRNA) or heart (protein)]	9.2	0.9	4.0×10−5	0.17
DQ597690	chr15:102304722–102304753 (telomere)	close to TM2D3 [brain (mRNA) or heart (protein)]	9.6	0.9	4.6×10−5	0.66
DQ597805	chr16:33732676–33732704	close to TP53TG3C [testis (mRNA and protein)]	13.1	0.9	4.8×10−5	0.37
DQ572493	chr19:51734482–51734513	intron 4 of CD33 [white blood cell (mRNA and protein)]	4.7	−0.9	8.7×10−5	0.02
DQ570746	chrM:295–323 (Mitochondrial telomere)	close to NADH1 [brain (mRNA and protein)]	16.4	0.9	1.9×10−5	0.33
DQ582201	chrM:296–324 (Mitochondrial telomere)	close to NADH1 [brain (mRNA and protein)]	16.6	0.9	5.5×10−5	0.62

r, correlation coefficient;

^*α=10⁻⁴

mRNA and protein expression of the Piwi protein in the prefrontal cortex:

piRNAs regulate cellular activities by interacting with Piwi proteins to form a complex ³⁴. The golden criterion to confirm the validity of piRNAs is the existence of Piwi proteins in the same tissue. To test the expression of Piwi in prefrontal cortex, the mRNA and protein expression of Piwi, including PIWIL1, PIWIL2, PIWIL3 and PIWIL4 was examined in the above four independent cohorts, including BRAINEAC, BioGPS, GTEx and ProteinDB.

Results

Ten piRNAs were significantly associated with years of survival:

Among the 9,453 brain piRNAs, 1,109 (11.7%) were nominally associated with years of survival (longevity) (p<0.05). With correction for multiple testing, associations for 10 piRNAs remained significant (4.0≤normalized log2-transformed intensity≤16.6; α=10⁻⁴; Table 1), including DQ581919 close to IQSEC1, DQ594437 close to GRM4, DQ575255 at intron 1 of CYP19A1, DQ597687 and DQ597690 close to TM2D3, DQ597805 close to TP53TG3C, DQ570746 and DQ582201 close to NADH1 (r=0.9), DQ593237 at intron 1 of CYP19A1 and DQ572493 at intron 4 of CD33 (r=−0.9) (9.9×10⁻⁵≤p≤1.9×10⁻⁵).

DQ597687 and DQ597690 are located in the telomeres of Chromosome 15, and DQ570746 and DQ582201 are located in the mitochondria. The potential TE targets of these piRNAs and their locations are listed in Table S3. These TEs include two LTRs, three LINEs and two SINEs. DQ593237 at CYP19A1 might target the simple repeat (GTCA)n, and no TE targets were found for the two mitochondrial piRNAs DQ570746 and DQ582201. All of the 10 piRNAs were not associated with AD (p>0.05), except for DQ572493 (p=0.02).

Multiple risk protein-coding genes and Piwi protein were expressed in the prefrontal cortex:

IQSEC1, GRM4, NADH1 and TM2D3 showed significant mRNA (36< normalized intensity <955 or 1<TPM≤31010) or protein (47ppm≤concentration≤198ppm) expression in the prefrontal cortex in four independent cohorts (Table S2). CD33 showed very low density of mRNA expression in one cohort, and CYP19A1 and TP53TG3C were not detected in the prefrontal cortex in any cohort. Specifically, in both UK European cohorts (BRAINEAC and BioGPS), IQSEC1, GRM4 and TM2D3 mRNAs were expressed in the prefrontal cortex (36.7≤ normalized intensity ≤954.4); CD33 showed very low density of mRNA expression in BRAINEAC cohort (normalized intensity=26.0); CYP19A1 was not detected, and TP53TG3C and NADH1 were not examined for brains in either cohort (Table S2). However, CYP19A1, TP53TG3C and CD33 were most abundant in the placenta (normalized intensity =4872), testis (normalized intensity =110) and whole blood (normalized intensity =86.8), respectively, in the BioGPS cohort (Table 1). Secondly, in the American cohort (GTEx), IQSEC1, NADH1 and TM2D3 mRNAs were abundantly expressed in the prefrontal cortex (41.4≤TPM≤31010.0). Other genes were not detected in the prefrontal cortex; however, CD33, CYP19A1 and TP53TG3C were most abundant in white blood cells (TPM=38.4), ovary (TPM=16.8) and testis (TPM=2.3), respectively (Table 1). (3) Finally, in the Germany cohort (ProteinDB), proteomic analysis showed that IQSEC1 and NADH1 proteins were abundantly expressed in the prefrontal cortex (47ppm≤concentration≤198ppm); CYP19A1, TM2D3, TP53TG3C and CD33 were not expressed in the prefrontal cortex, but most abundantly in the placenta (concentration=1084ppm), heart (19ppm), testis (7ppm) and white blood cells (12ppm), respectively (Table 1). Additionally, among the four Piwi proteins, PIWIL4 was detected in the prefrontal cortex in the American cohort (TPM=1.5) (Table S2).

Discussion

We previously reported that piRNAs were abundant in the prefrontal cortex ³. In this study, we found that ten brain piRNAs were significantly associated with years of survival. These ten piRNAs might probably play a potential role in determining lifespan. Among them, all seven intergenic and one intronic piRNAs were positively associated with age (correlation coefficient r>0), and two intronic piRNAs were negatively associated with age (r<0). Specifically, six intergenic piRNAs might probably suppress TE activity, regulating the transcription of IQSEC1, GRM4, CYP19A1, TM2D3 and TP53TG3C, respectively, maintaining the genomic integrity and slowing down the aging process ^14,35; one intronic piRNA, i.e., DQ572493, might probably suppress TE activity, activating the transcription of CD33 and accelerating the aging process; another intronic piRNA, i.e., DQ593237 at CYP19A1, and two other mitochondrial piRNAs, i.e., DQ570746 and DQ582201 close to NADH1, might probably accelerate aging by down-regulating the expression of the two protein-coding genes via TE–independent mechanisms. On the other hand, correlation analysis does not establish a causal relationship, which would require new studies employing cause and effect analyses.

IQSEC1, GRM4 and TM2D3 are most abundantly expressed in the prefrontal cortex. It is highly likely that the three genes’ product represent direct targets of the retrotransposons L2, LTR75 and AluJo, respectively, as controlled by the piRNAs. Aberrant proteostasis, a hallmark of aging ¹⁷, of IQSEC1 protein is a specific phenotype of brain aging in mammals and has been implicated in age-associated cognitive dysfunction ³⁶. GRM4 receptors substantially decrease with aging ^37,38, and have served as a treatment target of neurodegenerative conditions including Parkinson’s disease, AD, depression and schizophrenia ³⁹. TM2D3 protein has been implicated in apoptosis and regulation of cell cycle and repair ⁴⁰.

CD33 mRNA was specifically enriched in microglia and myeloid cells ⁴¹. DQ572493 at CD33 was the only piRNA associated with AD in this study. CD33 DNA variants have been associated with AD at genome-wide level (p<5e-8), a finding replicated in at least two genome-wide association studies ^42,43. Thus, CD33 is an AD-related myeloid protein. Additionally, CD33 conduces to the atrophy of AD-related brain structures and exacerbation of cognitive deficit, induces apoptosis, inhibits microglia-mediated clearance of A42, and reduces phagocytosis that restricts cell damages in the central nervous system ^41,44,45. Therefore, the increase of CD33 expression plays an important role in aging. piRNA DQ572493 might possibly suppress LTR14B activity, up-regulating CD33 expression and accelerating the aging process.

CYP19A1 encodes aromatase that is responsible for the aromatization of androgens into estrogens. Aromatase catalyzes the last step of estrogen biosynthesis from androgens. A CYP19A1 polymorphism has been associated with age at natural menopause ⁴⁶. Aromatase is a biological marker of the proliferation of human ovarian granule cells ⁴⁷. Suppression of aromatase may lead to premature ovarian aging in women ⁴⁷ and bone loss in men ⁴⁸. Although we did not find expression of CYP19A1 in our human brain samples, its brain expression was reported in other species ⁴⁹. Aromatase in the brain is usually expressed only in neurons. It has also been shown to decrease apoptosis following brain injury in zebra finches ⁵⁰. Suppression of aromatase would limit estrogen biosynthesis, reducing the neuroprotective actions of estrogens including estradiol, and thereby promoting apoptosis and aging ⁵¹. Estradiol may also regulate telomerase expression and activities that preserve telomeres, and slow down the aging process ⁵². piRNA DQ575255 might suppress retrotransposon L2b activity, up-regulating the transcription of CYP19A1, and promoting years of survival; conversely, piRNA DQ593237 may suppress CYP19A1 expression via TE-independent mechanism, e.g., regulating the expansion of the simple repeat (GTCA)n, and thus reduce lifespan.

NADH1, an mtDNA-encoded protein, is extremely abundant in the prefrontal cortex. It could contribute to years of survival by slowing down the actions of the electron transport chain, reducing free radical production ^53,54. NADH1 protein level in the brain significantly decreases with age ⁵⁵; it is widely implicated in aging and age-related diseases ⁵⁴. Two mitochondrial piRNAs might up-regulate NADH1 expression via TE-independent mechanism and thus promote years of survival.

Additionally, the two intergenic piRNAs DQ597687 and DQ597690 are located in telomeres, and might regulate telomere activity by direct sequence complementarity. Telomere shortening has been widely recognized as a hallmark of aging ¹⁷. Furthermore, mitochondrial dysfunction resulting from the alteration of NADH1 protein also represents a hallmark of aging ¹⁷.

Finally, we note that CYP19A1 and TP53TG3C are most abundant in the reproductive system. It is predictable that the brain piRNAs within or proximate to these two genes are also supposed to be expressed in this system, where they were initially discovered before incorporation into the microarray. piRNAs might probably regulate these genes’ expression in the reproductive system, with alterations of gene expressions leading to other age-related physiological dysfunctions that warrant investigations.

Half of the subjects in the present study had Alzheimer’s disease (AD). However, this phenotype was not associated with the expression of nine of the ten longevity-associated piRNAs, sex, age, or postmortem interval in our sample. DQ572493 at CD33 was the only piRNA that was modestly associated with AD (p=0.02), and disease status has been included as a covariate in the association analysis between years of survival and this piRNA expression. Therefore, the disease status did not represent a confounding factor in its significant association with years of survival.

We conclude that piRNAs might probably play an important role in determining the years of survival of humans. piRNAs might regulate aging-related gene expression to influence life span primarily via TE-dependent but also via TE–independent pathways. The hypothesized molecular mechanisms, including the regulation of piRNAs on mRNA or protein expression and the effects of gene expression on aging, remain to be investigated, for example, using genome editing technology for cause and effect analysis.