Neurotoxic mechanisms of DNA damage: focus on transcriptional inhibition

Abstract

Although DNA damage-induced neurotoxicity is implicated in various pathologies of the nervous system, its underlying mechanisms are not completely understood. Transcription is a DNA transaction that is highly active in the nervous system. In addition to its direct role in expression of the genetic information, transcription contributes to DNA damage detection and repair as well as chromatin organization including biogenesis of the nucleolus. Transcription is inhibited by DNA single strand breaks and DNA adducts. Hence, transcription inhibition may be an important contributor to the neurotoxic consequences of such types of DNA damage. This review discusses the existing evidence in support of the latter hypothesis. The presented literature suggests that neuronal DNA damage interferes with the RNA-Polymerase-2-dependent transcription of genes encoding proteins with critical functions in neurotransmission and intracellular signaling. The latter category includes ERK1/2 MAP kinase phosphatases whose lowered expression results in chronic activation of ERK1/2 and its reduced responsiveness to physiological stimuli. Conversely, DNA damage-induced inhibition of RNA-Polymerase-1 and the subsequent disruption of the nucleolus induce p53-mediated apoptosis of developing neurons. Finally, decreasing nucleolar transcription may link DNA damage to chronic neurodegeneration in adults.

1. Introduction

Currently, it is widely recognized that excessive DNA damage disrupts the proper function of the nervous system. Significant progress has been achieved in understanding the mechanisms of acute neurotoxicity of DNA damage including identification of ATM or PARP as DNA damage-activated inducers of neuronal death (Kauppinen & Swanson 2007, Katyal & McKinnon 2008). However, the DNA damage-associated mechanisms underlying chronic neuronal dysfunction and/or slowly progressing neurodegeneration are not yet identified. In this review, we discuss the existing evidence to support the contribution of transcriptional inhibition to both acute and chronic neurotoxicity of DNA damage.

2. Functions of transcription

Transcription is RNA synthesis on a DNA template, a process required for the expression of most genes (Fig. 1). Protein coding genes of the nuclear genome are transcribed into mRNAs by RNA Polymerase-2 (RNAPol2) (Lee & Young 2000). In contrast, genes for RNAs that do not encode proteins (often referred as the non-coding- or ncRNAs) are transcribed by the RNA Polymerases-1 (RNAPol1), 2 or 3 (RNAPol3). The ncRNA accounts for most of the RNA in cells. RNAPol1 transcribes 45S ribosomal RNA (pre-rRNA) which then is processed to mature forms of 5.8-, 18-, and 28S rRNA (Grummt 2003). The rRNA is the most abundant type of RNA. In addition, at least in cycling cells and in some tissues, RNAPol1 mediates the most transcriptional activity. The RNAPol1-driven transcription occurs in a nuclear subdomain known as the nucleolus that contains ribosomal DNA (rDNA) consisting of several hundred copies of the 45S rRNA genes (approximately 400 per haploid human genome). There is a growing number of ncRNAs that appear to be transcribed by RNAPol2. In fact, new methods of transcriptome analysis reveal that more than 90% of the human genome is transcribed, largely due to RNAPol2 activity (Birney et al. 2007). Although functional significance of most of these ncRNA transcripts is unclear, some of them give rise to small nuclear/nucleolar RNAs (sn/snoRNAs) and microRNAs (miRNAs) regulating RNA processing and translation/heterochromatin maintenance, respectively (Faller & Guo 2008, Egloff et al. 2008, Dieci et al. 2009). RNAPol3 transcribes several small RNAs including transfer RNAs (tRNAs) and 5S ribosomal RNA (White 2008). Finally transcription in mitochondria is carried out by the mitochondrial RNA polymerase POLRMT that is structurally unrelated to the eukaryotic nuclear RNA polymerases (Scarpulla 2008).

Figure 1.

Functions of the transcription. While gene expression is the most recognized role of the transcription, that process has also other functions which are not directly related to gene expression (see text for details).

Besides its direct role in gene expression, transcription also has other functions (Fig. 1). One of them is transcription-coupled DNA repair (TCR). The TCR is activated by “bulky” DNA adducts consisting of relatively large chemicals that are covalently bound to DNA. The bulky adducts cause transcriptional arrest blocking transcript elongation (Hanawalt & Spivak 2008). The transcriptional machinery scans the genome for such lesions and initiates their repair. While TCR has been most extensively characterized for RNAPol2-mediated transcription, it may also be active in RNAPol1-transcribed rDNA (Hanawalt & Spivak 2008). For both RNAPol1 and RNAPol2-transcribed genes, TCR may not only remove the bulky damage but also counteract gene silencing activity of DNA methyltransferases by removing methylated cytosines from the transcription-regulating regions (Barreto et al. 2007, Schmitz et al. 2009). Thus, DNA repair and epigenetic regulation of gene expression are dependent on ongoing transcription. Transcription may also serve as a genome integrity sensor initiating genotoxic stress response. For instance, it has been proposed that during transcript elongation DNA damage-triggered arrest of RNAPol2 activates apoptosis that is mediated by the DNA-damage regulated tumor suppressor p53 (Derheimer et al. 2007). Indeed, such a genome monitoring roles of the RNAPol2-mediated transcription fit well with observations that nearly the whole genome undergoes RNAPol2-dependent transcription (Birney et al. 2007).

Finally, nuclear organization and regulation of gene expression requires ongoing transcriptional activity. For instance, RNAPol1 activity maintains the structural integrity of the nucleolus (Rubbi & Milner 2003). In addition, in yeast, an initial round of RNAPol2-mediated transcription may remodel chromatin to enable the subsequent transcriptional activation of gene expression (Hirota et al. 2008). Anti-sense transcription that interferes with the transcriptional initiation of gene expression is another example of “expressionless” transcription that has an important regulatory role in the genome (Martens et al. 2004).

In mammalian brain, the portion of the mRNA-coding genome that is transcribed is 3–5 times greater than elsewhere in the body (Hahn & Laird 1971, Bantle & Hahn 1976). Also, robust rRNA synthesis has been detected in both developing and adult rat brain neurons and glia (Nievel & Kirby 1966, Lovtrup-Rein & Grahn 1970, Stoykova et al. 1979, Stoykova et al. 1983). Transcriptional inhibitors applied to various animal models have revealed the importance of transcription for the physiology of the nervous system including long-term synaptic plasticity, learning and memory, or completion of developmental cell death (Matthies 1989, Matthies et al. 1990, Oppenheim et al. 1990). In addition, as neuronal proteins and RNAs are turned over, transcription is required for maintenance of the nervous system (Finch & Morgan 1990). Consistent with these notions, transcriptional toxicity of the mutant huntingtin has been proposed as one of the major mechanisms of chronic neurodegeneration observed in Huntington's disease (Cha 2007). In addition, neuron-specific inactivation of the RNAPol2-associated transcription factor CREB or the general RNAPol1 co-activator TIF1A induced neurodegenerative diseases in mice (Mantamadiotis et al. 2002, Parlato et al. 2008). Hence, transcriptional deficiencies may serve as direct triggers for neurodegeneration and/or neuronal dysfunction. The neurotoxic consequences of transcriptional inhibition may be due to a loss of gene expression as well as a disruption of the gene expression-unrelated functions of transcription.

3. Neurotoxicity of DNA damage

The accumulation of DNA damage, including DNA single strand breaks (SSBs) and oxidative DNA adducts, is well documented in common age-related neurodegenerative disease and during normal aging (as reviewed extensively by Markesbery & Lovell 2006, Martin 2008, Yang et al. 2008, Yankner et al. 2008, Brasnjevic et al. 2008). Thus, it has been proposed that age related DNA damage may predispose to neuronal dysfunction culminating in neurodegenerative diseases such as Alzheimer's disease (AD), Parkinson's disease (PD) or ALS. Likewise, increased SSBs and oxidative DNA adducts have been proposed to contribute to neuronal loss following acute insults to the nervous system including stroke or traumatic brain injury (Nagayama et al. 2000, Liu 2001, Clark et al. 2001, Mendez et al. 2004, Kauppinen & Swanson 2007). Conversely, several DNA damaging chemicals that are used as anti-cancer agents are neurotoxic (for reviews see Kannarkat et al. 2007, Ahles & Saykin 2007). For instance, cisplatin and other related compounds can cause peripheral sensory neuropathy which is often a dose-limiting toxicity in clinical settings (Kannarkat et al. 2007). While there is a controversy as to whether the genotoxic anti-cancer drugs such as methotrexate induce permanent damage to the adult brain, they may cause persistent neurological deficits in children (Robaey et al. 2008).

The most informative insight into neurotoxic consequences of DNA damage comes from studies of inborn errors of DNA repair. Thus, Ataxia Telangectasia (AT), Ataxia with Oculomotor Apraxia 1 (AOA1), Spinocerebellar Ataxia with Axonal Neuropathy 1 (SCAN1) or Xeroderma Pigmentosum (XP) are caused by deficiencies in various DNA repair pathways and often involve neurodegeneration (recent extensive reviews on that subject include Rass et al. 2007, Katyal & McKinnon 2008, Caldecott 2008, and Brooks 2008). DNA double strand breaks (DSBs) or SSBs or bulky DNA adducts are expected to accumulate in AT or AOA1/SCAN1 or XP, respectively. These diverse forms of DNA lesions may activate non-overlapping genotoxic stress response pathways and engage distinct mechanisms of neurodegeneration. For instance, it has been proposed that the AT-related neurodegeneration is of developmental origin with its primary cause being reduced apoptosis of early post-mitotic neurons with chromosomal abnormalities (Katyal & McKinnon 2008). In contrast, neurodegeneration in AOA1, SCAN1 or XP is believed to originate from progressive, age-dependent accumulation of unrepaired DNA damage (Katyal & McKinnon 2008, Caldecott 2008, Brooks 2008). It has been suggested that such persistent DNA damage may induce neurodegeneration by reducing transcription of genes that are critical for neuronal maintenance (Katyal & McKinnon 2008, Caldecott 2008, Brooks 2008). As in the case of inborn DNA repair defects, also in spontaneous neurodegenerative diseases, different types of DNA damage may engage distinct mechanisms of neurotoxicity. Such mechanisms likely include transcriptional inhibition.

4. Transcriptional consequences of DNA damage

Similar to genotoxic interferences with DNA synthesis, DNA damage effects on transcription include lesion-induced arrest of transcription elongation or transcriptional bypass of unrepaired DNA lesions (for review, see Saxowsky & Doetsch 2006). Various types of DNA damage differ in their ability to induce one or the other outcome on transcription. Thus, bulky adducts, including the oxidation-generated 8,5′-(S)-cyclo-2′-deoxyadenosine, and SSBs trigger transcriptional arrest (Saxowsky & Doetsch 2006, Brooks 2008, Kathe et al. 2004). Conversely, small oxidative adducts such as the 8-oxo-7,8-dihydroguanine (8-oxoG) may be bypassed by the transcriptional machinery resulting in mutant transcripts (Saxowsky et al. 2008). In addition to interference with transcription elongation, DNA damage may also disrupt transcriptional initiation. For instance, oxidative damage to the promoter regions, including either small or bulky adducts, may reduce their activity by disrupting transcription factor binding (Ghosh & Mitchell 1999, Marietta et al. 2002). Finally, DNA damage-activated signaling may indirectly reduce transcription of undamaged chromatin. Thus, in cycling cells challenged with various DNA damaging agents, the DSB-activated protein kinase ATM inhibits RNAPol1 (Kruhlak et al. 2007). Likewise, RNAPol2 may undergo inhibitory phosphorylation and proteasome-mediated degradation in response to DNA damage (Saxowsky & Doetsch 2006).

As neurons are postmitotic cells, DNA damage interference with transcription may explain both acute and chronic neurotoxicity associated with at least some forms of genomic lesions (Katyal & McKinnon 2008, Brooks 2008). Alternatively, it has also been proposed that accumulation of unrepaired damage in non-coding DNA becomes a trigger for neuronal apoptosis upon an attempt to replicate neuronal genome (Nouspikel & Hanawalt 2003). Moreover, it has been suggested that, at least in relatively young neurons, the DSB toxicity is mediated by disruption of DNA replication that is initiated as a component of the damage response (Kruman et al. 2004, Barzilai et al. 2008). The cell cycle hypothesis of DNA damage neurotoxicity may explain the sensitivity of developing neurons to undergo DNA damage-induced apoptosis (Katyal & McKinnon 2008). However, it does not exclude other apoptosis-regulating mechanisms of sensing the genotoxic stress such as those involving transcriptional disruption (Rubbi & Milner 2003, Derheimer et al. 2007). Conversely, in chronic neurodegenerative diseases, DNA damage-induced neuronal death is of lesser significance as neurological deterioration correlates best with synapse loss but not cell death (Scheff & Price 2003, Gould et al. 2006). Thus, impairment of neuronal transcription appears as a reasonable candidate mechanism underlying chronic neurotoxicity of DNA damage. Although that concept is not new (for instance, see (Martin 1977) or (Finch & Morgan 1990)), there are relatively few published studies that directly investigated transcriptional consequences of neuronal DNA damage.

5. Reduced neuronal gene expression in response to aging-associated oxidative DNA damage

Aging is accompanied by the accumulation of oxidative DNA damage leading to increased risk of cancer and neurodegenerative diseases (for review, see Katyal & McKinnon 2008, Yankner et al. 2008, Brasnjevic et al. 2008). As little cell loss accompanies normal human brain aging, consequences of age-related DNA damage may include reduced transcription without acute cytotoxicity (Yankner et al. 2008, Brasnjevic et al. 2008).

In animals, direct transcriptional monitoring that relies on measuring the incorporation of labeled RNA precursors into newly synthesized RNAs has been used to study the transcriptional effects of aging. Following systemic administration of ¹⁴C-labeled uridine, a strong decline in transcription was reported in most brain regions of aged rats (MacKinnon et al. 1969). Similar changes were observed in isolated neuronal nuclei from aged rats whose transcriptional activity was analyzed by the run-on RNA synthesis assay (Lindholm 1986). The latter approach also suggested that transcriptional activity in glia nuclei is not affected by aging (Lindholm 1986).

More recently, measurements of RNA levels using RNA microarrays revealed a set of neuron-specific genes whose transcription may be reduced by DNA damage in aging human brain cortex. Thus, Lu, Yankner et al. analyzed cortical RNA levels of 30 individuals whose age ranged from 26 to 106 years (Lu et al. 2004). Comparing results from people younger than 43 vs. older than 72, 182 genes were identified that were downregulated more than 1.5 fold in the aged group. The most downregulated functional clusters of mRNA-coding genes included synaptic transmission, vesicular transport, Ca²⁺ handling/signaling, and MAP kinase/PKC signaling. These changes were unlikely due to neuronal loss as that is not observed during normal human aging (Yankner et al. 2008, Brasnjevic et al. 2008). Instead, the accumulation of unrepaired oxidative DNA damage has been demonstrated in several of the downregulated genes (Lu et al. 2004). The damage consisting of 8-oxoG adducts was promoter specific. High content of GC was proposed as a potential factor contributing to the promoter selectivity. It has also been proposed that the promoter activity of the down-regulated genes is more sensitive to 8-oxoG. While no mechanism for such hypersensitivity has been identified, it has been speculated that the oxidative adducts down-regulated some but not all genes by disrupting specific transcription factor-promoter interactions (Fig. 2). Conversely, absence of the TCR in the non-transcribed portions of the promoters has been proposed as a contributing factor to the apparent promoter preference for accumulation of unrepaired DNA damage (Lu et al. 2004). However, the actual reasons accounting for age-related accumulation of the promoter damage are to be identified.

Figure 2.

Hypothetical model of transcriptional inhibition observed in aging human cerebral cortex as proposed by Li, Yankner et al. (2004). Aging-associated accumulation of unrepaired oxidative damage to guanine bases (8-oxoG) disrupts interaction of the promoters with some transcription factors (such as a hypothetical TF-Y) reducing transcription initiation (see text for details).

In humans, the spectrum of mRNAs undergoing age-associated downregulation overlapped between various areas of the cortex (Fraser et al. 2005). The lack of similar changes in the cerebellum was proposed to be a result of relatively lower metabolic rates in that structure rendering cortex DNA more exposed to the oxidative damage (Fraser et al. 2005). Interestingly, the observed changes in cortical mRNAs appear human specific. While there is some overlap with the age-related mRNA declines in the cortex of the rhesus monkey, little similarity has been observed between aging humans and chimpanzees (Fraser et al. 2005, Loerch et al. 2008). Moreover, in mouse cerebral cortex, the most prominent transcriptome change consists of the up-regulation of mRNAs associated with an inflammatory/glia response (Loerch et al. 2008). Hence, the primate- but not the mouse cerebral cortex may respond to age-dependent DNA damage by inhibiting transcription. Alternatively, in rodents, transcriptional inhibition may have been undetected by RNA level analysis due to (i) increased RNA stability of the investigated transcripts, and/or (ii) omission of the prominent non-coding RNAs, such as the 45S rRNA transcript, and/or (iii) the stochastic nature of the transcription-interfering DNA damage resulting in no identifiable pattern of RNA changes. Indeed, earlier studies revealed decrease in nascent RNA production in aged rodent brains (MacKinnon et al. 1969; Lindholm 1986). However, as the sets of age-affected genes differ even between closely related primate species (Fraser et al. 2005), the neurotoxic “effector” mechanisms of age-related DNA damage are unlikely conserved during evolution.

While interpreting the results of RNA level measurements one has to realize that they provide only indirect indication of transcriptional activity. Thus, besides the rate of transcription, RNA levels are also affected by the rate of RNA decay (Houseley & Tollervey 2009). As the rate of RNA decay is often dependent on transcription, changes in RNA levels may underestimate differences in transcriptional activity. In addition, many popular platforms of large scale RNA analysis such as RNA microarrays are designed primarily to detect mature transcripts. Those are generally more stable than the primary unspliced transcripts and, as such, less sensitive to changes in transcription rates (Houseley & Tollervey 2009). Also, the published aging microarray studies used RNA obtained from homogenized brain tissue containing more glia than neurons. Conversely, under stress conditions, glia cell growth and proliferation are often increased implying enhanced transcription of many genes (Norenberg 1994). Hence, the transcriptional deficiency in the affected neurons may be masked by the reactive gliosis. Finally, at the cellular level, DNA damage may target genes in a stochastic manner (Vijg 2004). Therefore, analysis of the tissue RNA that represents many cells may identify only the secondary changes in gene expression rather than the primary target genes for DNA damage. Despite these obvious limitations, studies of RNA levels using microarrays provided so far the best evidence to support the notion that DNA damage-related transcriptional inhibition contributes to neuronal dysfunction in humans.

Taken together, there is a reduction of several neuronal RNAPol2-transcribed mRNAs in cerebral cortex of aging human brain. However, the mechanisms underlying most of the observed decreases remain to be identified. Finally, the functional impact of the identified transcriptional defects has not yet been evaluated.

6. Impaired inactivation of ERK1/2 signaling as a consequence of DNA damage-induced transcriptional inhibition

The Extracellular signal Regulated Kinases-1/2 (ERK1/2), also known as the mitogen-activated protein kinases-2/1 (MAPK2/1), are the founding members of the MAP kinase family (Pearson et al. 2001). ERK1/2 is activated by phosphorylation at the Thr-183/Tyr-185 residues. This phosphorylation is mediated by the dual specificity Ser/Thr and Tyr kinase MAP kinase kinase-1/2 (MKK1/2, also known as MEK1/2 or MAP2K1/2) and opposed by several dual specificity- and/or Tyr- and/or Ser/Thr-specific phosphatases (Saxena & Mustelin 2000, Kondoh & Nishida 2007) (Fig. 3). Conversely, MKK1/2 is activated through phosphorylation by the Ser/Thr protein kinase Raf. ERK1/2 is highly expressed in neurons providing a signal transduction unit that rapidly responds to extracellular signals. In neurons those signals include neurotrophins, neurotransmitters and/or electrical activity. ERK1/2 couples such signals to changes in neuronal excitability and/or long lasting modifications of neuronal gene expression contributing a critical switch for survival, morphogenesis and synaptic plasticity (Adams & Sweatt 2002, Kaplan & Miller 2000, Miller & Kaplan 2003, Hetman & Gozdz 2004, Parrish et al. 2007). As ERK1/2 is an activator of gene expression, genes of several phosphatases that deactivate ERK1/2 are induced in the ERK1/2-dependent manner providing a negative feedback loop to terminate ERK1/2 signaling (Kang & Kim 2006, Ekerot et al. 2008) (Fig. 3).

Figure 3.

Hypothetical model explaining DNA damage-associated activation of the neuronal ERK1/2 signaling pathway. In neurons, synaptic activity-induced increase of cytosolic [Ca²⁺] or neurotrophin-mediated activation of tyrosine receptor kinases (Trks) initiate the signaling by the kinase cascade of Raf-MKK1/2-ERK1/2. Thus, ERK1/2 is activated following its phosphorylation by MKK1/2. Conversely, its inactivation is carried out by phosphatases of phospho-ERK1/2 (PPE). The targets of activated ERK1/2 include transcriptional regulators that induce several PPEs terminating ERK1/2 signaling. DNA damage disrupts ERK inactivation by reducing transcription of PPEs (see text for details).

Cisplatin is an anti-cancer drug that induces intra- and inter-strand crosslinks in the DNA (Trimmer & Essigmann 1999). The latter type of damage potently blocks transcription and is believed to be critical for the cytotoxicity of cisplatin (Cullinane et al. 1999). Cisplatin has well-recognized neurotoxic effects on the peripheral nervous system that appear related to its DNA damaging potential (McDonald et al. 2005, Kannarkat et al. 2007, Dzagnidze et al. 2007). In various types of cells, including human cancer cells or cultured rat primary neurons, cisplatin induces delayed and long lasting activation of the ERK1/2 pathway (Jo et al. 2005, Gozdz et al. 2003, Brozovic & Osmak 2007). For instance, in cultured cortical neurons from newborn rats, accumulation of activated ERK1/2 starts at six hr and keeps rising at least for up to twelve hr after adding cisplatin (Gozdz et al. 2003). This is in contrast to ERK responses to the neurotrophin BDNF or the neurotransmitter glutamate which occur within minutes following stimulation (Hetman et al. 1999, Bading & Greenberg 1991).

As the ERK1/2 is activated by phosphorylation, ERK1/2 activation can also be a result of reduced dephosphorylation. Such a scenario could fit well with the delayed character of the ERK response to cisplatin suggesting slow accumulation of the activated ERK1/2. Indeed, in cisplatin-treated neurons, the rate of ERK1/2 dephosphorylation decreased raising a possibility that cisplatin-induced DNA damage disrupted the expression of transcriptionally-regulated ERK1/2 phosphatases (Gozdz et al. 2008). Coinciding with the ERK1/2 activation, cisplatin reduced incorporation of the RNA precursor 5-fluorouridine (5-FU) into nascent RNA all over the nucleus suggesting general transcriptional inhibition (Gozdz et al. 2008). That was accompanied by reduced levels of mRNAs for the major neuronal ERK phosphatases including DUSP6 and the VRK3 subunit of DUSP3/VHR. Importantly, similar to cisplatin, a general transcriptional inhibitor, Actinomycin D (ActD), delayed dephosphorylation of neuronal ERK1/2 while triggering delayed accumulation of activated ERK1/2 (Gozdz et al. 2008). Finally, in cisplatin- or ActD-treated neurons, ERK was relatively more affected than the JNK MAP kinase pathway (Gozdz et al. 2008). Hence, accumulation of activated ERK1/2 has been identified as a response to transcriptional inhibition following neuronal DNA damage.

A long lasting accumulation of phosphorylated ERK1/2 has been observed in degenerating neurons in Alzheimer's and Parkinsons's diseases as well as their animal and/or cell culture models (Zhu et al. 2004, Chu et al. 2004, Stein & Johnson 2002, Chu et al. 2004, Zhu et al. 2004, Dagda et al. 2008). Reduced activity of ERK1/2 phosphatases has been proposed as the primary cause underlying chronic activation of neuronal ERK1/2 in these pathologies (Chu et al. 2004). Interestingly, declining cortical mRNA levels of several ERK phosphatases, including DUSP3/4/5/6 and PTPN5/STEP, were detected in aged human brains (Loerch et al. 2008). Hence, in the aging human cerebral cortex, ERK inactivation may be negatively affected by the DNA damage interference with transcription of the ERK phosphatases. Moreover, such transcriptional impairment may synergize with direct oxidative damage of ERK phosphatases to produce the persistent accumulation of activated ERK1/2 (Foley et al. 2004; Levinthal & DeFranco 2005).

An important question is that of the consequences of reduced ERK1/2 dephosphorylation to neuronal ERK1/2 signaling and neuronal functions. Previously reported kinetical models of the ERK1/2 signaling unit indicated the importance of various negative feedback mechanisms, including the ERK1/2 phosphatases, in determining the duration of ERK1/2 signaling (Bhalla et al., 2002; Orton et al. 2005). Conversely, in a neural cell line, the timing of ERK1/2 signaling is a critical factor for the stimulus specificity of ERK1/2-mediated responses (Sasagawa et al. 2005, Santos et al. 2007). Hence, one could predict that upon dysfunction of the negative feedbacks, those stimuli that elicit cellular responses through relatively transient ERK activation will become uncoupled from their physiological effectors. Moreover, as ERK activation by physiological neuronal activity is fairly brief in duration, the impaired negative feedback of the neuronal ERK unit would likely prevent ERK-dependent responses to neuronal activity, such as synaptic plasticity or memory formation (Adams & Sweatt 2002; Ajay & Bhalla 2004). In addition, an experimentally-validated modeling study suggested that activity of ERK1/2 phosphatases is critical for the responsiveness of the ERK signaling unit to tightly timed stimuli (Bhalla et al., 2002). Thus, we used a simplified computational model of the core ERK pathway, including the Raf-MKK-ERK kinase cascade as well as the MKK-, and ERK phosphatases, to investigate potential consequences of DNA damage-associated impairment of ERK dephosphorylation on ERK signaling (Fig. 4A). Assuming constant activities of Raf, an ERK core activator, and, of the MKK/ERK phosphatases that oppose it, the levels of pERK increase to equilibrium with the non-phospho-ERK1/2 (Fig. 4B). If the model includes an exponential decline of ERK1/2 dephosphorylation activity overlapping with that observed in cisplatin-treated neurons (three-fold decline of the Tyr-185 dephosphorylation rate constant after twelve hr cisplatin exposure, (Gozdz et al. 2008)), pERK levels increase until nearly the entire pool of non-phospho-ERK1/2 is converted to pERK1/2 by 30 hr (Fig. 4C). Hence, inability to activate ERK in response to new physiological stimuli will be the obvious consequence of impaired dephosphorylation of ERK1/2. Another consequence will be the disrupted encoding of temporal information due to loss of the timing control over ERK signaling.

Figure 4.

The kinetical model of dysregulated ERK signaling following DNA damage-associated loss of ERK phosphatases. A, The differential equation model presented in top panel is used to describe the longitudinal relation between Raf (r), MKK1/2 (x), ERK1/2 (y), and PPE (z). Dotted quantities indicate derivatives with respect to time. The kinetic constants were picked so as to approximate the published experimental data (Gozdz et al. 2008; α₁=0.2, α₂=0.1, β₁=0.2, β₂=0.1, γ₁=0.1, γ₂=0.2). The additional constants A (=10) and B (=5) represent the amounts of MKK and ERK at the inception of the signaling process; at that time, the amounts of Raf and PPE were assumed to be 0.1 and 3, respectively. B, C, Trajectories of the differential equation representing the changes in the amounts of RAF (r), MKK (x), ERK (y), and PPE (z) over 60 hour period of time. In B, we take δ=0, i.e., assume that PPE does not decrease in time from its initial amount. In C, we take δ=0.0916 and thus allow PPE to decrease in time. The rate of the decrease is fitted to be consistent with experimental results indicating a three-fold decrease in PPE activity for phospho-Tyr185-ERK1/2 after Treating cultured rat cortical neurons with 10 μg/ml cisplatin for 12 hr (marked as the black point on the PPE graph, (Gozdz et al. 2008)). With constant PPE activity, ERK and phospho-ERK levels remain at equilibrium (B). Upon PPE decline, all ERK is converted to phospho-ERK suggesting loss of ERK responsiveness to physiological stimuli and, therefore, functional inactivation of this signaling pathway (C).

Besides preventing physiological ERK signaling, the prolonged accumulation of active ERK1/2 may produce additional “gain of function” effects including non-apoptotic neuronal death or the cytoskeletal dysfunction in response to excessive phosphorylation of the microtubule-associated protein tau (Gozdz et al. 2003, Stanciu & DeFranco 2002; Mazanetz & Fischer 2007). Overall, in neurons that are challenged with DNA damage, transcriptional inhibition of ERK1/2 phosphatases likely contributes to the dysfunction of the ERK1/2 signaling pathway. However, it remains to be tested whether a similar mechanism operates in neurodegenerative diseases.

7. Nucleolar stress as a neurotoxic consequence of DNA damage

The nucleolus is a structure within the nucleus that contains several hundred clustered repeats of 45S rRNA genes, also known as ribosomal DNA (rDNA) (Grummt 2003, Drygin et al. 2010). The RNAPol1-mediated nucleolar transcription of the 45S rRNA accounts for at least 50% of the total transcriptional output of a cell and initiates the nucleolus-based process of ribosomal biogenesis. As the ongoing nucleolar transcription is required for maintenance of the nucleolus, this structure rapidly disintegrates following RNAPol1 inhibition (Rubbi & Milner 2003, Yuan et al. 2005). Recent studies suggest that in both yeast and mammalian cells, 45S rRNA transcription is the dominant force modulating ribosomal biogenesis. Thus, the constitutively active form of the RNAPol1 co-activator transcription initiation factor-1A (TIF1A) not only boosts rDNA transcription but also elevates ribosomal protein- and 5S rRNA levels, increases the number of ribosomes and stimulates protein synthesis (Laferte et al. 2006).

Ribosomal biogenesis is the best characterized nucleolar function with the most robust representation in the nucleolar proteome (Andersen et al. 2005). However, this structure also has other emerging roles including processing and assembly of non-ribosomal ribonucleoprotein complexes such as U6 splicosome and miRNAs (Patel & Bellini 2008, Shiohama et al. 2007). In addition, the nucleolus serves as a site for sequestration of various regulatory proteins. For instance, in cycling cells, nucleolar disassembly following DNA damage-induced RNAPol1 inhibition results in release of p53 activators such as ribosomal proteins L5, L11 and L23, or the tumor suppressor p14ARF leading to p53-dependent apoptosis (Rubbi & Milner 2003, Olson 2004a). Such observations led to a concept that disruption of nucleolar transcription produces a state of nucleolar stress that in turn is a trigger for p53-dependent apoptotic cell death.

Several features of RNAPol1-driven transcription make it a good candidate for a genome integrity checkpoint. The RNAPol1-driven transcription provides a major contribution to the total transcriptional output of the cell (Grummt 2003). The rate of rDNA transcription is reduced by various forms of DNA damage and recovers after the damage is repaired (Kruhlak et al. 2007, Zhang et al. 1988). The 45S pre-rRNA is encoded by similarly organized and regulated rDNA loci whose high copy number and constitutively high transcription rate provide a consistent indicator of genomic integrity (Grummt 2003). Lastly, as rRNA is expressed in both terminally postmitotic and actively dividing cells, the nucleolar transcription checkpoint may be independent of the ability to enter the cell cycle (Grummt 2003).

In neurons, nucleolus was first observed in the early 19^th century (Olson 2004b). In fact, histological studies of these cells led Valentin to propose the term nucleolus in 1839. Metabolic labeling studies confirmed that neuronal nucleoli contain active RNAPol1 (Nievel & Kirby 1966, Lovtrup-Rein & Grahn 1970, Stoykova et al. 1979, Stoykova et al. 1983). The possibility that neuronal nucleolar transcription serves as a sensor of DNA damage is supported by observations of nucleolar dysfunction/disorganization in DRG neurons from rodents treated with DNA damaging anti-cancer drugs including cisplatin, other Pt derivatives and the DNA topoisomerase-1 (Topo1) inhibitor doxorubicin (Tomiwa et al. 1986, Bigotte & Olsson 1987, McKeage et al. 2001, Jamieson et al. 2005). The reduced neuronal nucleolar volume indicative of RNAPol1 inhibition has been reported at the early stages of Alzheimer's disease and during normal rodent aging (Mann et al. 1988, Garcia Moreno et al. 1997). In a mouse model of accelerated aging caused by the loss of the klotho gene transcript, oxidative neuronal DNA damage, reduced RNAPol1 activity and neurodegeneration were also reported (Nagai et al. 2003, Anamizu et al. 2005). Finally, hypoxia/ischemia induces nucleolar disruption in brain neurons (Mosgoeller et al. 2000, Kastner et al. 2003, Kerr et al. 2007). As AD, aging and hypoxia are associated with the accumulation of DNA damage, these results provide further, albeit indirect, support for the concept that nucleolar transcription senses neuronal DNA damage.

The role of nucleolar stress as a switch for the p53-dependent neuronal apoptosis in response to SSB DNA damage was probed in cultured cortical neurons from newborn rats (Kalita et al. 2008). These cells die in an apoptotic manner if challenged with the SSB-inducing Topo1 inhibitor camptothecin (CPT) (Morris & Geller 1996, Hetman et al. 1999). Moreover, p53 is required for CPT-induced neuronal apoptosis (Xiang et al. 1998, Kalita et al. 2008). Using an in situ run on assay with 5-FU, it was demonstrated that nucleolar production of nascent RNA was blocked as early as 1 hr following CPT addition to the neuronal cultures (Kalita et al. 2008). At the same time, extranucleolar labeling of nascent RNA was not disturbed by CPT. In CPT-treated neurons, the expression level ratio of the relatively unstable pre-rRNA transcript (45S rRNA) to its relatively stable mature derivative, the 18S rRNA, was also decreased. Finally, the preferentially nucleolar localization of the chaperone protein B23/nucleophosmin whose nucleolar residence is strictly dependent on RNAPol1 activity was disrupted by CPT. All these data support a conclusion that in cultured neurons, CPT-induced SSBs blocked nucleolar transcription. In addition, CPT triggered nucleolar transcription inhibition even if CPT-induced apoptosis was prevented by overexpression of the dominant-negative mutant form of p53, indicating that nucleolar stress may serve as a signal activating p53-dependent apoptosis in response to DNA damage. Indeed, selective disruption of RNAPol1 transcription by knocking down the RNAPol1 co-activator TIF1A was sufficient to induce the p53-dependent neuronal apoptosis. Finally, that apoptotic response required de-novo expression of protein coding genes. Altogether, it appears that SSBs induce apoptosis of developing cortical neurons by inhibiting nucleolar transcription. The resulting nucleolar stress leads to p53-dependent apoptosis (Fig. 5). As that apoptotic response requires expression of protein coding genes, one could speculate that the ability of transcription-interfering DNA damaging agents to induce neuronal apoptosis will depend on their relatively greater inhibitory effects on nucleolar- than the extranucleolar transcription.

Figure 5.

Hypothetical model of nucleolar involvement in neuronal response to DNA damage. DNA damage blocks the RNAPol1-mediated transcription of rDNA disrupting structural integrity of the nucleolus. Thus, functions of that nuclear domain are compromised leading to disinhibition of p53, disturbed processing of various non-rRNA species and reduced ribosomal biogenesis. In developing neurons that are challenged with DNA damage, such a nucleolar stress may activate the p53-dependent apoptotic program whose execution requires induction of the RNAPol2-transcribed killer genes (Kalita et al. 2008). In mature neurons, which are more resistant to apoptosis, chronic consequences of nucleolar dysfunction may include atrophy, loss of synapses/neurites and reduced synaptic plasticity (see text for details).

The aforementioned sufficiency of nucleolar stress to induce neuronal apoptosis is also supported by in vivo studies using a neuron-specific mouse knockout of the floxed TIF1A alleles (Parlato et al. 2008). Crossing these animals with the nestin-driven CRE recombinase mice resulted in TIF1A deletion restricted to neuronal precursors. Such deficiency triggered p53 activation and massive apoptosis of these cells resulting in an extensive loss of brain tissue in newborns.

In addition to inducing SSBs, CPT also inhibits Topo1 activity (Champoux 2001). That enzyme has been suggested to play a crucial role in RNAPol1 transcription elongation by removing the transcription-induced torsion of rDNA (Ljungman & Hanawalt 1996). Hence, nucleolar transcription could be inhibited by reduced Topo1 activity rather than the SSB DNA damage. However, our recent data support the notion that neuronal DNA damage is sufficient to block nucleolar transcription (M. Pietrzak, S.C. Smith, C. Gomes and M. Hetman, in preparation). In cultured cortical neurons, nucleolar disruption was observed following treatment with the DNA topoisomerase-2 (Topo2) inhibitor etoposide that triggers a mixed SSB-/DSB DNA damage. In contrast, Topo2 inhibition using the non-DNA damaging inhibitor ICRF-193 affected neither nucleolar transcription nor nucleolar structure (M. Pietrzak, S.C. Smith, C. Gomes and M. Hetman, in preparation).

It is important to note that not all types of DNA damage may require nucleolar stress to induce neuronal apoptosis. Such universal coupling between the DNA damage and the apoptosis has been proposed in cycling cells (Rubbi & Milner 2003). However, our recent results from cortical neurons indicate that the selective DSB rDNA damage by the overexpressed restriction endonuclease I-Ppo-1 (Berkovich et al. 2007) is a powerful trigger of p53-dependent apoptosis without any detectable effect on nucleolar transcription and/or nucleolar integrity (M. Pietrzak, S.C. Smith, C. Gomes and M. Hetman, in preparation). Thus, at least in neurons that were acutely challenged with DNA damage, SSBs but not DSBs induce nucleolar stress. Conversely, DSBs induce a signaling pathway converging with nucleolar stress to activate the pro-apoptotic p53. The anti-nucleolar potential of other forms of DNA damage, including oxidative adducts, remains to be determined.

SSBs that in neurons induce nucleolar stress are the type of DNA damage that is observed in neurodegenerative diseases (Brasnjevic et al. 2008). In SSB-challenged neurons, nucleolar disruption persists even after preventing apoptosis (Kalita et al. 2008). Therefore, nucleolar stress may also contribute to the non-apoptotic, chronic consequences of neuronal DNA damage (Fig. 5). Such effects, including reduced protein synthesis, neuronal atrophy, or synaptic loss, may follow nucleolar stress-associated decreases in ribosomal biogenesis. Indeed such a scenario has been proposed based on morphological studies of degenerating motoneurons in the klotho-deficient mice (Anamizu et al. 2005). Alternatively, the non-ribosomal effects of nucleolar stress such as U6/miRNA processing defects may also contribute to neurodegeneration. Finally, the nucleolarly-produced rRNA has been suggested to act as a major cellular anti-oxidant that controls the intracellular iron homeostasis (Honda et al. 2005). It remains to be tested whether in neurons, chronic nucleolar disruption leads to translational deficits and/or widespread defects in gene expression and/or oxidative stress.

Additional support for a possibility that DNA damage-induced nucleolar stress contributes to chronic neurodegeneration comes from the studies of the transgenic mice whose RNA-Pol1 activity was blocked in adult forebrain neurons by the tamoxifen-induced deletion of the floxed TIF1A alleles (Parlato et al. 2008). In these animals, loss of hippocampal neurons was observed at 3- but not 1 month after the tamoxifen treatments. Therefore, in adult neurons, slowly progressing neurodegeneration but not an immediate apoptosis may be a direct consequence of nucleolar dysfunction. Finally, a recent study in severely depressed humans with history of childhood abuse established a correlation between the hypermethylation of rDNA promoters, lowered nucleolar transcription and hippocampal atrophy (McGowan et al. 2008).

Of note, in D.melanogaster, transient induction of nucleolar DSBs resulted in rDNA recombination, loss of heterochromatin and reduced number of the 45S rRNA gene repeats (Paredes & Maggert 2009a, Paredes & Maggert 2009b). Therefore, although DSBs do not appear to trigger acute nucleolar stress, they may reduce nucleolar functions in a chronic and decremental manner. Hence, it is possible that such DSB-associated conditions as AT involve rDNA loss and reduction in nucleolar transcription as one of the mechanisms underlying neurodegeneration. It remains to be tested whether DSBs or other forms of DNA damage contributed to the rDNA loss reported in the brains of aged dogs (Johnson & Strehler 1972).

Overall, nucleolar transcription appears to be an important target for neuronal DNA damage. RNAPol1 inhibition induces nucleolar stress disrupting various functions of that nuclear domain including ribosomal biogenesis or inhibition of the pro-apoptotic activity of p53. The interesting possibility that nucleolar stress contributes to neurodegenerative diseases requires further evaluation.

8. Concluding remarks and future perspectives

Transcriptional inhibition occurs in response to various forms of DNA damage including the neurodegeneration associated SSBs, oxidative adducts and interstrand crosslinks. In neurons or neuronal cell lines the transcription inhibitory potential of each of these lesions has been demonstrated. So far dysfunction of the neuronal ERK1/2 signaling pathway and activation of p53-dependent neuronal apoptosis have been identified as the consequences of DNA damage-associated transcriptional interference.

The emerging role of transcriptional inhibition in neurotoxicity of DNA damage raises several important issues that should be addressed in the future. First, conclusive evaluation of the possibility that DNA damage-induced transcriptional inhibition is a major neurotoxic mechanism in common neurodegenerative diseases requires the development of techniques allowing direct evaluation of transcriptional activity across the neuronal and/or glial cell genome, preferably at the single cell level. Such technology would enable testing a hypothesis that transcription of specific gene clusters, for instance, the nucleolar rDNA, is hypersensitive to neurodegeneration-associated DNA damage. Also, the transcriptional inhibitory potential and transcription inhibition mechanisms of various types of neuronal DNA damage are yet to be determined. Importantly, one could consider that at least some negative consequences of DNA damage on transcription could be reduced to decelerate neurodegeneration. Thus, in the case of the oxidative promoter damage one could propose counteracting its hypoactivity with such drugs as histone deacetylase inhibitors that would enhance gene expression by modifications of the epigenetic environment of the injured promoters (Langley et al. 2005). Conversely, as every cell appears to have a pool of epigenetically silenced 45S rRNA genes (McStay & Grummt 2008), nucleolar disruption could be potentially attenuated by mobilization of such silent genes. Another interesting, yet untested, possibility is that DNA damage-induced transcriptional inhibition helps to protect the neuronal genome against further damage. Finally, as DNA damage-associated loss of mitochondrial transcription has been proposed to contribute to heart disease (Tsutsui et al. 2009), it is tempting to speculate that a similar mechanism is involved in neurodegeneration. Obviously, addressing these issues will fill important gaps in the understanding of the nervous system response to DNA damage. Conversely, DNA damage-mediated transcriptional inhibition may turn out to be a suitable target for anti-neurodegenerative interventions.

Abbreviations

5-FU

5-fluorouridine

8-oxoG

8-oxo-7,8-dihydroguanine

ActD

Actinomycin D

AD

Alzheimer's disease

ALS

amyotrophic lateral sclerosis

AOA1

ataxia with oculomotor apraxia 1

AT

ataxia telangiectasia

ATM

ataxia telangiectasia mutated

BDNF

brain-derived neurotrophic factor

cisplatin

cis-platinum(II) diamine dichloride

CPT

camptothecin

DSB

DNA double strand break

DUSP

dual specificity phosphatase

ERK1/2

extracellular signal regulated kinase1/2

JNK

c-jun N-terminal kinase

MAPK1/2

mitogen-activated protein kinases-1/2

MKK1/2

MAP kinase kinase-1/2

ncRNA

non-coding RNA

PARP

poly(ADP-ribose)polymerase

PD

Parkinson's disease

PPE

phosphatase of phospho-ERK1/2

rDNA

ribosomal DNA

RNAPol

RNA-polymerase

SCAN1

spinocerebellar ataxia with axonal neuropathy 1

SSB

DNA single strand break

TCR

transcription-coupled DNA repair

TIF1A

transcription initiation factor-1A

Topo1

DNA-topoisomerase-1

XP

xeroderma pigmentosum