Loosening chromatin and dysregulated transcription: a perspective on cryptic transcription during mammalian aging

Abstract

Cryptic transcription, the initiation of transcription from non-promoter regions within a gene body, is a type of transcriptional dysregulation that occurs throughout eukaryotes. In mammals, cryptic transcription is normally repressed at the level of chromatin, and this process is increased upon perturbation of complexes that increase intragenic histone H3 lysine 4 methylation or decrease intragenic H3 lysine 36 methylation, DNA methylation, or nucleosome occupancy. Significantly, similar changes to chromatin structure occur during aging, and, indeed, recent work indicates that cryptic transcription is elevated during aging in mammalian stem cells. Although increased cryptic transcription is known to promote aging in yeast, whether elevated cryptic transcription also contributes to mammalian aging is unclear. There is ample evidence that perturbations known to increase cryptic transcription are deleterious in embryonic and adult stem cells, and in some cases phenocopy certain aging phenotypes. Furthermore, an increase in cryptic transcription requires or impedes pathways that are known to have reduced function during aging, potentially exacerbating other aging phenotypes. Thus, we propose that increased cryptic transcription contributes to mammalian stem cell aging.

Aging is a complex, multifaceted process that involves changes in many aspects of cell biology, from genomic instability to metabolism to cell–cell signaling [1]. However, what causes these alterations and what effects they have on the cell and the organism are not as well understood. An altered chromatin structure, one of the hallmarks of aging, has the potential to impact many molecular aspects of aging through its regulation of transcription. Indeed, the more open chromatin structure in aged cells and organisms [2] has been linked to several types of transcriptional dysregulation, including reactivation of retrotransposons [3, 4] and activation of the senescence-associated secretory phenotype [5]. In yeast and worms, intragenic cryptic transcription, a type of transcriptional dysregulation in which aberrant transcription is initiated from non-promoter regions, increases during aging; in yeast, this is caused by a reduction in trimethylated lysine 36 of histone H3 (H3K36me3) across gene bodies and the resulting disruption to chromatin structure [6]. Whether elevated cryptic transcription occurs during mammalian aging, or contributes to this process, is only beginning to be explored.

In this review, we discuss the possibility that cryptic transcription increases during mammalian aging, specifically in stem cells, which are thought to play an outsized role in this process [7], and that elevated cryptic transcription contributes to aging phenotypes. We first delve into the molecular mechanisms that regulate this phenomenon, which has been described in both embryonic and adult stem cells derived from mice and humans. Evidence that genetic perturbations that promote a chromatin state permissive for cryptic transcription disrupt stem and progenitor cell function is detailed, and the ways in which increased cryptic transcription may impact cellular function are described. We compare known chromatin and cellular changes that occur during aging to what is known about the biology governing cryptic transcription, highlighting the similarities between the two, and discuss a recent work that links altered chromatin structure in aged stem cells with increased cryptic transcription. Finally, we explore how aging phenotypes may both contribute to an increase in cryptic transcription and exacerbate the deleterious effects of this phenomenon.

Open chromatin in the wake of RNA polymerase II promotes intragenic cryptic transcription

Intragenic cryptic transcription is a phenomenon that has been studied for yeast for nearly 20 years [8]. During this process, RNA polymerase II (Pol II) aberrantly initiates transcription from promoter-like sequences within the gene body [9]. In this organism, mutations that prevent the reversal of chromatin state changes that occur during transcription to allow Pol II progression along the DNA increase cryptic transcription [9–11]. More recently, it was shown that perturbations that alter the post-transcription chromatin state also promote cryptic transcription in mammals [11–14]. During Pol II transit, the nucleosome structure is disrupted, resulting in a more open chromatin state immediately following Pol II transcription [15]. There is also co-transcriptional conferral of H3K4 methylation and H3K36me3, driven by the association of MLL1 and SETD2, respectively, with elongating Pol II [16, 17]. Not all of these changes persist, however, and the gene bodies of actively transcribed genes are maintained in a relatively closed chromatin environment. This ‘resetting’ largely depends upon H3K36me3, which recruits the FACT complex, KDM5B and DNMT3B to restore a closed chromatin conformation after transcription. The FACT complex functions as a chaperone to restore nucleosomes in the wake of Pol II transit [15], while KDM5B removes co-transcriptionally conferred H3K4me3 from the gene body [12]. Rather than reversing transcription-associated chromatin changes, DNMT3B deposits de novo CpG methylation intragenically [14]. Thus, although transcription elongation opens the chromatin, additional co-transcriptional histone modifications function to assemble various complexes to restore a closed chromatin conformation after Pol II transit (Figure 1A).

Figure 1 .

Regulation and molecular effects of cryptic transcription in mammals. (A) The chromatin landscape of an actively transcribed gene. H3K36me3 is co-transcriptionally conferred by SETD2 and acts as a scaffold to recruit KDM5B and DMNT3B. These enzymes, in turn, remove H3K4me3 and catalyze DNA methylation, respectively, which function to repress transcription from intragenic cryptic promoters. CpG methylation of DNA is denoted by the red pentagons and the (inactive) cryptic promoter by the dashed gray arrow; full-length transcripts are shown above the gene. (B) As chromatin structure becomes disrupted with age, or by loss of any of the noted enzymes, the chromatin state of transcribed genes is more open and permissive for RNA polymerase II (Pol II) entry, with decreased H3K36me3 and DNA methylation and increased H3K4me3 in the gene body. The polymerase can then activate transcription from cryptic promoters, generating truncated mRNA molecules, shown as the shorter transcripts above the gene. Panels (A) and (B) are based on work published in refs. [12–14]. (C) The consequences of increased cryptic transcription. Association of initiating Pol II with cryptic promoters may inhibit the progression of Pol II from the endogenous promoter, a phenomenon known as transcriptional interference. This can disrupt transcriptional networks and gene expression. Cryptic transcripts themselves are subject to one of two fates. They can be degraded by the RNA exosome (1) or processed as endogenous mRNAs, exported from the nucleus, and translated by the ribosome (2). The peptides synthesized from cryptic transcripts can likewise go down different pathways. Some may properly fold into active proteins and go on to function in the cell (3); such proteins may lack regulatory or enzymatic domains, or encode an entirely different peptide than the endogenous transcript if the cryptic transcript caused a frameshift. The peptides may also misfold (4), engaging the chaperone system, and possibly titrating factors needed for the proper folding of endogenous proteins. Finally, the aberrant peptides generated from cryptic transcripts may be degraded by the proteasome (5). The cellular effects of increased cryptic transcription, transcripts and their protein products are based on refs. [6, 14, 41, 44].

When this process is disrupted and the chromatin of actively transcribed genes is not restored to a closed state, intragenic cryptic transcription increases (Figure 1B). Loss of SETD2, which confers H3K36me3, elevates cryptic transcription in human mesenchymal stem cells (MSCs) [13] and murine embryonic stem cells (ESCs) [14]. In ESCs, this loss of Setd2 prevents intragenic localization of Dnmt3b; knockout of Dnmt3b in this system drives cryptic transcription independently of Setd2 [14], though it should be noted that other groups have found that inducible double knockout of Dnmt3a and Dnmt3b; growth of ECS in conditions that inhibit DNA methylation; and inducible triple knock out of Dnmt1, Dnmt3a and Dnmt3b do not cause an increase in cryptic transcription [18, 19]. This discrepancy could be due to inducible versus constitutive reduction in DNA methylation or the redistribution of DNA methylation that occurs when different DNA methyltransferases are perturbed [20]. Also in ESCs, the loss of Kdm5b, which is recruited to transcribed genes by H3K36me3, causes an increase in intragenic cryptic transcription [12]. In each of these cases, it is easy to understand how the resulting altered chromatin structure could promote the entry of Pol II at intragenic promoter-like sequences. CpG methylation is associated with promoter silencing; the reduction of this modification due to Dnmt3b loss would thus inhibit the silencing of potential cryptic promoters within the gene. Likewise, increased intragenic H3K4me3 caused by Kdm5b loss promotes an active promoter-like chromatin state within the gene body. As both these enzymes are localized to chromatin by H3K36me3, conferred co-transcriptionally by SETD2, loss of SETD2 would cause the same downstream changes as loss of Dnmt3b or Kdm5b, along with potentially maintaining the chromatin in a relatively nucleosome-depleted state, due to reduced FACT recruitment. Thus, as in yeast, a failure to reset the chromatin along gene bodies after Pol II transit results in increased cryptic transcription in mammalian stem cells.

Perturbations that elevate cryptic transcription reduce stem and progenitor cell function

The fact that multiple mechanisms are used to prevent intragenic cryptic transcription suggests that elevated cryptic transcription is deleterious. The deep conservation, from yeast through mammals, of co-transcriptionally conferred H3K36me3 as a scaffold to assemble factors that maintain a closed chromatin state that inhibits aberrant transcription initiation is also indicative of this. Indeed, in yeast, increased cryptic transcription both decreases replicative lifespan and sensitizes this organism to caffeine and rapamycin, which mimic nutrient stress by dysregulating TOR signaling [6, 21]. In mammals, where cryptic transcription has not been as well studied, less is known, though there is emerging evidence that perturbations that increase cryptic transcription are harmful. While only a few studies have assayed for cryptic transcription or examined intragenic H3K4me3 or DNA methylation, knockout or knockdown of SETD2, KDM5B and de novo DNA methyltransferases negatively impacts stem and progenitor cell function. We focus on these cells because (1) whole body knockouts of these genes are lethal in mice [22] and (2) stem cells may have an outsized role in aging [7].

Loss of Kdm5b in ESCs inhibits the self-renewal capacity of these cells, which was linked to increased cryptic transcription and intragenic accumulation of H3K4me3 in Kdm5b knockout cells [12]. Another group found that depletion of Kdm5b both reduces the self-renewal capacity of ESCs and disrupts embryoid body differentiation and teratoma formation, with particular defects in formation of mesoderm and ectoderm derivatives [23]. This group additionally showed that Kdm5b depletion allows the spreading of H3K4 methylation into the gene body, and found that genes that accumulate intragenic H3K4 methylation are more prone to transcriptional dysregulation [24]. In contrast, it has also been found that loss of Kdm5b does not impact ESC self-renewal, but prevents neural differentiation of these cells [25]. In line with the first two ESC studies, loss of Kdm5b limits the self-renewal capacity of hematopoietic stem cells (HSCs) [26]. Thus, while only a single study has directly linked increased intragenic cryptic transcription to Kdm5b loss and reduced stem cell function, there is broad consensus that Kdm5b function is essential for both ESCs and HSCs.

The role of DNMT3B, and DNA methylation, is complicated by the partially redundant de novo methyltransferase DNMT3A. Both genes are expressed in HSCs and ESCs, and both enzymes are recruited to chromatin at least in part by H3K36me3 [27, 28], though only DNMT3B has been directly implicated in repressing intragenic cryptic transcription [14]. The effects of loss of either or both de novo DNA methyltransferases in ESCs are controversial. There is no indication that such loss impairs the self-renewal of these cells [29–31]. Several groups have suggested a role for either or both DNMT3A and DNMT3B in neural differentiation, though their loss has been implicated in either promoting or inhibiting this process in different studies [30–33]. In contrast to ESCs, de novo DNA methylation is well known to be essential for HSC function. Loss of Dnmt3a or double knock out of Dnmt3a and Dnmt3b promotes HSC self-renewal at the expense of differentiation [20, 34]. As in ESCs, DNA methylation-based control of self-renewal versus differentiation gene networks was implicated as the role of DNMTs in HSCs. Overall, while DNMT3B and its paralog DNMT3A have essential roles in stem cells, it is unclear if the inhibition of cryptic transcription contributes to these roles.

Although SETD2 has been extensively studied for its role in the regulation of alternative splicing [35], less work has been done on its role in stem cell biology. While SETD2 knockdown does increase intragenic cryptic transcription in human MSCs, whether this perturbation effects their self-renewal or differentiation has not been characterized [13, 36]. Similarly, in their experiments assessing cryptic transcription in murine ESCs, Neri et al. [14] did not determine whether Setd2 loss affects the biology of these cells. Another report found that Setd2 knockout in murine ESCs does not affect self-renewal, strongly impairs endoderm differentiation, and weakly inhibits ectoderm differentiation while increasing cryptic transcription [37]. In HSCs, several groups have found that Setd2 loss both impairs self-renewal and skews differentiation potential toward erythroid cells [38, 39], though cryptic transcription was not assessed in this context. Interestingly, SETD5, a SETD2 paralog that confers H3K36me3 on transcribed genes in neural stem cells (NSCs) and the nervous system, is required for the normal proliferation of progenitor cells and for the correct wiring of the nervous system [40]. Additionally, in SETD5 haploinsufficient cells, an accumulation of nascent transcripts within gene bodies on DRB treatment was observed [40], similar to the accumulation of RNA Pol II along gene bodies in ESCs that lack Dnmt3b that is suggestive of increased cryptic transcription [14]. Thus, SETD2 and its paralog SETD5 in the nervous system play critical roles in inhibiting cryptic transcription and stem and progenitor cell biology.

Cryptic transcripts and functional consequences in cells

While there is a strong correlation between perturbations that increase cryptic transcription and deleterious outcomes, such an association could result from other characterized functions of these perturbations on gene regulation. Several hypotheses, with varying levels of experimental support, explain how cryptic transcription may interfere with normal cellular processes. These include inhibiting transcription from endogenous promoters; titrating transcription factors and RNA Pol II away from endogenous promoters, and thus altering transcriptional networks; and imposing an energy drain on the cell by inappropriately engaging the transcriptional and translation machinery [6]. The ORFs encoded by cryptic transcripts may also encode proteins that differ from the ORFs of their cognate endogenous transcripts; if translated, the resulting proteins may lack functional or regulatory domains or, if a frameshift has occurred, be entirely unique [6, 41]. Directly testing these hypotheses is a nontrivial endeavor, so while it is known that increased cryptic transcription is deleterious in yeast and mammals, we do not have a complete understanding of why this is so. Nevertheless, there is some direct evidence of the molecular impact of increased cryptic transcription on cells (summarized in Figure 1C).

Both sense and antisense transcripts that overlap transcription start sites (TSSs) can inhibit transcription from the overlapping TSS; this phenomenon is referred to as transcription interference [42]. It was recently shown in yeast lacking Setd2 that antisense intragenic cryptic transcripts can extend through the 5′ end of the gene from which they originate; furthermore, genes containing these cryptic transcripts showed reduced expression from the endogenous TSS, indicative of transcription interference [21]. While the effect of cryptic transcription on ‘normal’ transcription has not been directly studied in mammals, several lines of evidence suggest it might interfere with transcription from endogenous promoters. The GARP gene in humans is expressed from two alternative promoters; in most tissues, it is exclusively expressed from its upstream TSS, but in regulatory T cells, T cell receptor activation transiently drives expression from the downstream promoter. Activation of the downstream TSS limits transcription from the upstream promoter [43]. A similar phenomenon may occur when expression from cryptic TSSs increases. Additionally, in Setd5 haploinsufficient murine NSCs that show evidence of increased cryptic transcription [14, 40], the RNA polymerase II elongation rate is decreased [40], which could suggest that intragenic cryptic transcription interferes with elongation of endogenous transcripts.

While some cryptic transcripts are degraded by the RNA exosome complex [14, 41], there is ample evidence that other cryptic transcripts are stable, capped, polyadenylated molecules that mimic endogenous mRNAs, which a priori suggests they are capable of being translated. Sequencing technologies designed to detect these transcripts, such as TL-seq, DECAP-seq, PRO-cap and CAGE, isolate capped, polyadenylated RNAs from the steady-state pool of cellular RNA [14, 44–46]. Early work in yeast elegantly demonstrated that cryptic transcripts generate truncated proteins, relative to the cognate full-length transcripts [41]. Likewise, ribosome profiling of cryptic transcripts in wildtype yeast indicates that at least some of these molecules are associated with polysomes and the polysome signature around the putative start codons is similar to that observed at the start codons of endogenous transcripts [44]. In murine ESCs lacking Dnmt3b, ribosome profiling revealed that, globally, ribosomes are depleted from 5’ UTRs and accumulate within transcripts, particularly in retained introns, which suggests that the cryptic transcripts in these cells may be translated into aberrant proteins [14]. Thus, increased cryptic transcription can interfere with multiple cellular processes by interfering with ‘normal’ transcription; engaging the RNA exosome and ribosomes; and ultimately producing aberrant proteins that could either disrupt cellular function themselves or be targeted for degradation.

Cryptic transcription and mammalian aging

As discussed above, the underlying driver of cryptic transcription is an open chromatin state, with reduced H3K36me3 and DNA methylation and increased H3K4me3, that is permissive for RNA polymerase II entry. During aging, chromatin undergoes numerous changes, and altered chromatin structure is one of the now-canonical hallmarks of aging [1]. In mammals, these changes include histone/nucleosome loss, global loss of CpG methylation and a reduction in H3K36me3 levels [2]. Thus, the aged chromatin state mirrors that which promotes cryptic transcription, indicating that the chromatin in aged cells and organisms is more permissive for cryptic transcription than that in young animals. Indeed, a recent study has shown increased cryptic transcription in aging MSCs, HSCs and NSCs. Significantly, in both MSCs and HSCs, this elevation of cryptic transcription is associated with decreased H3K36me3 and CpG methylation and increased H3K4me3. Furthermore, in MSCs, initiating Pol II accumulates along gene bodies with age, and sites of cryptic initiation are associated with increased TATA binding protein, indicative of an increase in cryptic intragenic transcription initiation [47]. Together, these data suggest that the more open and permissive chromatin structure that develops during aging in mammalian stem cells promotes aberrant intragenic Pol II entry and causes an increase in cryptic transcription.

There is some evidence that elevated cryptic transcription has a direct impact on stem cell aging, in addition to the general disruption of function described above. Loss of Setd2 or Dnmt3a and Dnmt3b in HSCs phenocopies certain aspects of normal aging in these cells: conditional knockout of Setd2 in the hematopoietic lineage results in a slight myeloid differentiation bias [39], while double knock out of Dmnt3a and Dnmt3b promotes self-renewal at the expense of differentiation [20, 34], though the role of cryptic transcription in these phenotypes has not been assessed. Furthermore, several indirect lines of evidence suggest that increased cryptic transcription may impact mammalian aging. A number of molecular pathways that intersect with cryptic transcription and cryptic transcripts are impacted during aging, suggesting that these processes contribute to each other. For example, transcriptional dysregulation is a hallmark of aging [1] and increased cryptic transcription interferes with normal transcription [21], which could directly contribute to other changes in the transcriptome. Additionally, given that some cryptic transcripts are subject to RNA exosome-mediated decay [14], elevated cryptic transcription could congest this pathway and slow the removal of other transcripts targeted for degradation. Likewise, as cryptic transcription increases, more cryptic transcripts can associate with ribosomes; in principle, this could inhibit the translation of full-length transcripts and contribute to the age-associated loss of proteostasis [1]. It is also possible that translation of aberrant proteins from cryptic transcripts could exacerbate the loss of proteostasis. If such peptides are degraded, they could also contribute to the load on the proteosome and autophagy, and indeed, as these processes have reduced functionality during aging [1], this could in turn potentiate deleterious effects of cryptic transcripts.

Taken together, the studies discussed in this section provide compelling evidence that age-associated chromatin changes promote cryptic transcription in adult stem cells and that increased cryptic transcription may negatively impact both stem cell biology and function. Indeed, given that many of the cellular pathways with which cryptic transcripts are known to interact have reduced functionality in aged organisms, even baseline levels of cryptic transcription may have a greater negative effect upon the cell in the aged environment. This, coupled with the deleterious increase in cryptic transcription with age in other organisms and the fact that perturbations that increase cryptic transcription phenocopy some aspects of HSC aging, suggests that elevated cryptic transcription may play an underappreciated role in mammalian stem cell aging.

Concluding remarks

The study of cryptic transcription as it relates to mammalian aging is an emerging field, and many important questions remain to be answered. It will be critical to gain a more thorough understanding of the prevalence and tissue distribution of increased cryptic transcription with age, though there is some evidence that it may occur in a broad swath of tissues [47]. As the techniques that precisely map the 5′ ends of transcripts, and are thus most sensitive at detecting cryptic transcription, require large samples at the outset, further developing these technologies to accept smaller input samples will be necessary to robustly address this question. Another gap in our knowledge is how the open chromatin state that is permissive for cryptic transcription develops during aging; examining how the expression and genome-wide distribution of factors that regulate this process change with age is the first step to address this question. Already, there is tantalizing evidence that in several tissues, DNMT3B expression is reduced with age, which likely contributes to a reduction in DNA methylation [48–51]. However, much more work remains to be done in this area. Finally, it will be important to determine whether increased cryptic transcription contributes to aging phenotypes or is merely a consequence of the aging process. There is some evidence already that perturbations expected to increase cryptic transcription in HSCs recapitulate certain aspects of aging in these cells [20, 34, 38]. Whether this holds true in other tissues, and whether interventions that decrease cryptic transcription can mitigate aging phenotypes in aged animals or cells remains to be explored. Answering these questions will enhance our appreciation of the role of increased cryptic transcription in mammalian aging.

Key Points

• Open chromatin promotes cryptic transcription

• Increased cryptic transcription impairs stem cell function

• Cryptic transcription is found in aged stem cells

Brenna McCauley received her bachelor’s degree from Rice University and doctorate from Carnegie Mellon University. She is currently a postdoctoral research in the Huffington Center on Aging at Baylor College of Medicine. She was a recipient of NIH T32 training grant and is currently supported by funding from NIH and Welch Foundation.

Weiwei Dang received his bachelor’s degree from Peking University and doctorate from Southern Illinois University. He is currently assistant professor and CPRIT Scholar for Cancer Research in the Huffington Center on Aging at Baylor College of Medicine. His laboratory has been supported by multiple grant awards from NIH, and foundations, including Ted Nash Long Life Foundation and Welch Foundation.

Funding sources

The authors were supported by National Institutes of Health grant R01AG052507; and Welch Foundation grant Q-1986-20,190,330 to W.D.

Conflict of interest

The authors declare no conflicts of interest.