6.1. Huntington’s Disease
The majority of studies investigating the roles for PGC-1α in neurodegeneration revolve around determining its roles as an etiological factor and/or potential therapeutic target in Huntington’s Disease (HD) and Parkinson’s Disease (PD; see list of studies summarized in
Supplementary Table S1). HD is a severe progressive neurodegenerative disorder caused by a CAG repeat expansion in exon 1 of the huntingtin (
HTT) gene [
90]. Patients experience involuntary movements, abnormal gait and posture, and psychiatric symptoms, accompanied by cortical atrophy and the loss of SPNs of the caudate-putamen [
91,
92]. Mitochondrial dysfunction has been identified as a key contributor to HD pathology [
93,
94,
95,
96], with a reduction in mitochondrial complex activity being a reproducible finding in HD patient samples [
97,
98,
99,
100]. Murine models using quinolinic [
101] or 3-nitropropionic acid [
102] model striatal vulnerability and behavioral traits of HD by inducing excitotoxicity or directly targeting the mitochondria, respectively. Mutant huntingtin (mtHTT) itself can interfere with mitochondrial respiration, calcium buffering capacity, and ATP production [
103,
104]; array studies across multiple HD models have repeatedly identified a reduction in mitochondrial and synaptic transcripts in the striatum [
105,
106].
Various mechanisms of transcriptional dysregulation have been reported in HD [
107,
108,
109,
110,
111], with studies indicating the disruption of PGC-1α-dependent and PGC-1α-independent transcriptional programs. Initial interest in a link between PGC-1α dysfunction and HD etiology was stimulated by the observation of ambulatory hyperactivity and striatal vacuolizations in the PGC-1α null mouse and reduced viability of SPNs cultured from these mice [
46]. Follow-up studies found reductions in
PPARGC1A mRNA expression in HD patient brains and cell culture and mouse models of HD [
112,
113,
114,
115,
116,
117,
118] and suggested that mtHTT interferes with CREB-mediated transcription of
Ppargc1a [
112]. Importantly, these studies demonstrated that PGC-1α overexpression attenuates mtHtt-induced cellular toxicity [
112,
113]. Around the same time, another study documented a role for CREB in the regulation of PGC-1α in neurons in response to oxidative stress [
52]. The inhibition of PGC-1α expression by mtHTT has been proposed to contribute to NMDAR-mediated excitotoxicity [
117]. Interestingly, multiple studies documented disruption in PGC-1α-dependent transcriptional pathways in peripheral tissues as well, including brown adipose tissue [
113], muscle [
119], and liver [
120]; this suggests the ability of mtHTT to interfere with PGC-1α-dependent transcriptional programs in different cellular contexts.
Subsequently, several studies found an association between variation the
PPARGC1A locus and age of onset in HD [
121,
122,
123,
124], with the age-of-onset-associated haplotype encompassing the promoter of brain-specific isoforms [
88,
125]. However, it is important to point out that an association between age-of-onset and the
rs7665116 SNP in
PPARGC1A was not replicated when phenotypic and genotypic stratification was taken into account [
126]. Of note,
NRF1 and
TFAM are genetic modifiers of HD [
127], suggesting convergence of genetic risk on gene programs for mitochondrial function.
In addition to evidence for the disruption of PGC-1α expression and/or activity in HD, there is abundant evidence for the direct interaction of mtHTT with transcription factors [
128,
129,
130], causing a disruption in gene expression and compromising cellular viability [
108,
131,
132]. For example, mtHTT can prevent the transcriptional activator SP1 from binding to DNA to elicit transcription [
133,
134,
135] (reviewed in [
131]). SP1 can serve as both a mediator of PGC-1α interactions with other transcription factors and as a regulator of PGC-1α transcription itself [
136]. mtHTT is also capable of interacting with CBP to reduce its expression [
137,
138,
139,
140] and with CREB [
141], interfering in its interactions with TAFII130 and the activation of downstream genes [
142], including PGC-1α.
mtHTT can also interact with p53 [
137,
143], a robust transcriptional regulator of multiple mitochondrial programs in neurons [
144], increasing its expression and leading to mitochondrial dysfunction and reduced cell viability [
143]. Absence of p53 ameliorates these deficits and prevents the onset of behaviors seen in HD models [
143]. PGC-1α is reported to bind to and modulate p53 to promote cell survival, metabolic, and antioxidant transcriptional programs [
145,
146]. Heat shock transcription factor HSF1 could also be involved, as its binding at the PGC-1α promoter is significantly reduced in a cell culture model of HD and silencing of HSF1 in this model exacerbates cell death [
127].
These studies led us to test whether a reduction in PGC-1α in SPNs is sufficient to cause transcriptional changes, ambulatory hyperactivity, and SPN loss. While we found that deletion of PGC-1α specifically in SPNs causes age-related ambulatory hypoactivity and a reduction in previously identified PGC-1α-dependent genes such as
Nefh and
Idh3a, it does not cause striatal atrophy [
62]. Further, the expression of PGC-1α mRNA, as well as many of its dependent genes identified in other cell-types, was unchanged in the striatum of the HDQ [
62] and R6/2 [
45] mouse models. Considering the observations of changes in other models, it is possible that assessment of cell-type-specific transcriptional profiles (
Drd1 versus
Drd2-expressing SPNs) may reveal changes consistent with past studies. Overall, though, these findings suggest that disruptions in PGC-1α expression and/or activity are not sufficient to cause an HD-like phenotype in mice, in the absence of mtHTT. Considering that the expression of mtHTT selectively in striatal SPNs fails to recapitulate motor deficits seen in animals with global expression [
147], it is possible that PGC-1α deficiency is necessary in other cell types (striatal interneurons or cortical-striatal projection neurons) to generate an HD-like phenotype.
In fact, directed expression of mtHTT specifically in PV-expressing neurons causes ambulatory hyperactivity and cortical excitability [
148]. Being that PV-expressing neurons depend heavily on PGC-1α, this model in particular may be best to understand not only how the circuit is disrupted in HD but the role of PGC-1α in regulating it. Aside from the more obvious cortico-striatal pathway affected in HD, it would be remiss to exclude the possibility that a reduction in PGC-1α in the cerebellum could impact disease progression. Several studies in HD models indicate a disruption in Purkinje cell firing rate and eventual cell loss as well as reductions in
Pvalb mRNA expression [
149,
150]. In fact, Purkinje cells in the PGC-1α null mouse exhibit a reduction in firing rate and reductions in metabolic and synaptic transcripts and eventual cell loss [
77], indicating an eventual need for studies of this transcriptional pathway in models of HD.
A compelling demonstration of the global downregulation of genes involved in oxidative phosphorylation in SPNs in human HD and mouse HD models was recently published [
151]. In this study, the authors used single-nucleus RNA-sequencing and translating ribosome affinity purification to transcriptionally profile different neuron and glial types from human postmortem HD brain and HD mouse models. While genes involved in oxidative phosphorylation were reduced in both DRD1/
Drd1 and DRD2/
Drd2-expressing SPNs, the latter were more robustly affected in mouse models. While no changes in PGC-1α expression were reported, the authors speculate that disruption of RARβ (
Rarb), which is highly expressed in SPNs, could be involved [
152]. Of note, PGC-1α can interact with members of the retinoid receptor family [
153], so it is possible that some of the neuroprotective effects of PGC-1α overexpression in the striatum could be mediated through RARβ activation. It is also possible that splice variants of PGC-1α are decreased in these models; studies in both human postmortem tissue and HD models have reported a decrease in the N-terminal variant of PGC-1α with no effect on the full-length protein [
118]. Further studies of the cell-type-specific regulation of these variants and the mechanisms underlying their regulation are warranted, particularly in disease settings.
How these transcriptional changes could give rise to a relatively selective loss of SPNs is unclear. A combination of cell- and non-cell-autonomous mechanisms are proposed to contribute to SPN vulnerability in HD (reviewed in [
154]), some of which are related to cellular identity [
155]. SPNs are characterized by significantly long projections that may render them more vulnerable to disruptions in mitochondrial trafficking and clearance. Further, striatal mitochondria, relative to those in other cell types, are particularly sensitive to fluctuations in calcium concentrations and, as a result, produce lower levels of ATP than other neuron types [
156]. Additional potential reasons for SPN vulnerability include the interplay between GABA metabolism and the citric acid cycle; brain mitochondria have been shown to have lower content of succinate dehydrogenase, which could favor the maintenance of GABA levels over the conversion of succinate to fumarate [
156,
157,
158]. However, not all GABAergic populations are vulnerable; striatal PV-INs are lost with advancing disease in HD patients [
159] and in the R6/2 model [
160] while cholinergic interneurons and interneurons expressing somatostatin and/or neuronal nitric oxide synthase are relatively preserved [
161,
162,
163,
164].
Considering the evidence for disruption in multiple transcriptional pathways in HD patients and cell and animal models, it is likely that any intervention will need to engage multiple targets. While recent efforts to knock down the expression of the
HTT gene itself with antisense oligonucleotide therapy have shown promise [
165,
166,
167], approaches are still needed to complement this strategy (see therapeutic strategy section, below).
6.2. Parkinson’s Disease
In contrast to HD which is an autosomal dominant disorder with high penetrance, PD is thought to arise from a contribution of genetic and environmental factors. Recent estimates of heritability in PD range from 16–36% [
168], with the majority of mechanistic studies focusing on understanding molecular and cellular functions of rare genetic variants. Key clinical features of the disease include resting tremor, rigidity, and akinesia. Motor symptoms are accompanied by a progressive loss of dopaminergic neurons of the substantia nigra pars compacta (SNc) and, typically, the abnormal aggregation of the synaptic protein α-synuclein in deposits called Lewy bodies and Lewy neurites [
169,
170,
171].
Although the exact mechanisms of dopaminergic cell loss are unclear, abundant evidence suggests the involvement of mitochondrial dysfunction [
172,
173,
174]. Dopaminergic neurons of the SNc are especially vulnerable to mitochondrial toxins such as rotenone [
173], paraquat [
175], 6-hydroxydopamine [
176] and MPTP [
177,
178,
179]. A number of genes associated with autosomal dominant forms of PD including leucine-rich repeat kinase 2 (
LRRK2), vacuolar protein sorting-associated protein (
VPS35), and
SNCA have been linked to mitochondrial dysfunction. Additionally, three genes associated with autosomal recessive forms of parkinsonism,
PARKIN (PARK2),
PINK1, and
DJ-1 (PARK7), have roles in mitochondrial clearance in response to mitochondrial depolarization and oxidative stress (reviewed in [
180]). The strongest evidence for a mechanistic link between PD and the dysregulation of transcriptional programs for nuclear-encoded mitochondrial genes stems from two lines of investigation: identification of the mechanisms underlying neuronal vulnerability with PARKIN loss-of-function [
181,
182,
183,
184,
185] and the transcriptional profiling of dopaminergic neurons from postmortem tissue of patients with Lewy pathology [
186].
The protein α-synuclein, encoded by the PD-linked gene,
SNCA, is a major component of Lewy bodies which have recently been shown to also contain mitochondria, along with other organelles [
187]. Point mutations (A53T and A30P) in the
SNCA gene (
PARK1 locus) and triplication of the
SNCA locus are associated with upregulation and subsequent increase in α -synuclein expression [
188,
189,
190]. Studies from human iPSCs with A53T
SNCA mutations identify a MEF2C/PGC-1α transcriptional pathway contributing to neuronal damage. In these neurons, MEF2C and PGC-1α were downregulated through oxidative stress that inhibits MEF2C’s ability to regulate PGC-1α which prevents its neuroprotective effects [
191]. Similarly, in animals harboring the A30P mutation, PGC-1α was found to be downregulated, interfering with the transcription of neuroprotective genes [
192].
LRRK2 mutations cause a form of PD which is clinically indistinguishable from idiopathic PD [
193]. The most frequent mutation of LRRK2 (G2019S) causes an increase in its kinase activity, but it is unclear how this leads to PD. LRRK2 has been shown to be linked to mitochondrial dysfunction through regulation of mitochondrial motility; it has been shown to work in concert with PARKIN and PINK1 to modulate mitophagy [
194,
195]. Studies indicate that PGC-1α and the NAD-dependent protein deacetylase SIRT1 play a central role in cell metabolism and mitochondrial biogenesis [
196,
197]; deacetylation of PGC-1α by SIRT1 causes its activation and the transcription of genes involved in antioxidant defense [
198,
199,
200]. Interestingly, iPSC-derived dopaminergic neuron cultures, as compared to glutamatergic cultures from LRRK2 G2019S patients, have diminished expression of the active PGC-1α and a subsequent increase in ROS [
201]. In fact, a small-molecule activator of PGC-1α enhanced resistance against oxidative stress in human dopaminergic neurons [
202].
Mutations in the
PARKIN gene (~130 different mutations documented in ~1000 patients) cause early-onset PD with a median age of onset at 31 years of age [
203,
204]. PARKIN is an E3 ubiquitin ligase that plays a role in regulation of mitochondrial quality control via ubiquitination of toxic substrates for degradation by the proteasome [
205,
206]. Thus, PARKIN deficiency caused by loss-of-function mutations causes the accumulation of noxious substrates, leading to cellular stress [
207]. One of these substrates is PARKIN-interacting substrate (PARIS/
ZNF746/
Zfp746), which accumulates in the absence of PARKIN expression. The accumulation of PARIS interferes with the expression of PGC-1α and NRF-1 via an insulin response element in the
Ppargc1a promoter [
181], mimicking reductions in
PPARGC1A and
NRF1 mRNA expression in postmortem SNc from PD patients. Follow-up studies confirmed these findings in drosophila [
183] and stem cell models [
185], with the observation that PARIS-mediated dopaminergic vulnerability can also be rescued by overexpression of PINK1 [
183].
Transcriptional profiling studies support the idea that PGC-1α-dependent pathways of nuclear-encoded mitochondrial genes are impaired in PD [
186]. Laser capture microdissection of dopaminergic neurons from postmortem tissue of patients with Lewy pathology revealed a reduction in a number of PGC-1α-responsive genes, especially genes encoding proteins for respiratory complexes of the electron transport chain. While this study did not report reduction in
PPARGC1A mRNA or PGC-1α protein, these data suggest the disruption of a transcriptional program for oxidative phosphorylation in these neurons in PD, potentially prior to cell loss.
Consistent with these data, a reduction in PGC-1α expression in multiple brain regions, including the caudate-putamen, is observed in both patients with advanced stage PD and animal models of PD [
181,
192]. Interestingly,
PPARGC1A polymorphisms have been reported to influence the age of PD onset [
208]. Two studies have indicated that full-length brain-specific splice variants of PGC-1α are reduced in postmortem SNc of PD patients [
74,
209]. In one of the studies, this reduction was accompanied by an increase in splice variants encoding truncated isoforms which can interfere with the activity of the full length variant [
209]. Additionally, a loss of PGC-1α significantly exacerbates cell death in MPTP [
52,
210,
211,
212,
213,
214] and α-synuclein-induced cell death [
192,
215,
216] models of PD, while overexpression of PGC-1α enhances autophagy and reduces rotenone- and α-synuclein-mediated toxicity in cell culture [
186,
192,
217]. PGC-1α overexpression [
52,
212,
218,
219,
220,
221,
222] or stabilization [
211] can also prevent MPTP-mediated neurotoxicity in vivo. Recently, it has also been discovered that ferulic acid can reinstate mitochondrial dynamics through PGC-1α expression modulation in 6-hydroxydopamine lesioned rats [
223].
Despite the fact that
TFAM expression is not reduced in PD postmortem tissue [
181], multiple lines of evidence suggest that dopaminergic neurons are vulnerable to disruptions in mitochondrial DNA (mtDNA) maintenance and/or regulation, potentially independent of PGC-1α. mtDNA maintenance depends on several nuclear encoded genes that form the replisome for mtDNA replication. These genes are mtDNA polymerase gamma 1 (
POLG1),
TFAM (discussed above), and the DNA helicase TWINKLE (
TWNK). Mutations in these genes increase the risk of PD [
224]. POLG1, like other polymerases, controls mtDNA synthesis, repair and replication [
225]; the length of a CAG repeat in exon 2 directly correlates with function of the protein, and increased number of repeats has been associated with PD [
226,
227]. Also, mice expressing a proof-reading-deficient version of POLG1 on a PARKIN-deficient background exhibit an increase in pathogenic mtDNA mutations, motor deficits, and dopaminergic cell loss [
228].
There have also been mutations identified that lead to reduction in mtDNA copy number in PD patients [
229]. Accordingly, deletion of
Tfam in dopaminergic neurons of mice (the Mito-Park model) causes a depletion of mtDNA and age-related loss of dopaminergic neurons of the SNc [
230], and mice expressing mutant forms of
Twnk selectively in dopaminergic neurons exhibit an increase in mtDNA mutations and dopaminergic neuron degeneration with a Parkinson’s-like motor phenotype [
231]. Additionally, immunoprecipitation analyses have revealed that PARKIN itself can directly bind to mtDNA and TFAM, suggesting that PARKIN could directly affect the mitochondrial genome independent of PGC-1α [
232,
233,
234].
Considering the relatively low basal expression of
Ppargc1a and
Tfam mRNA in dopaminergic neurons in mice (
Figure 1a), it is possible that any stressors which reduce or impair the expression or function of these genes will overwhelm the ability of mitochondria to compensate, leading to cell loss. This relative deficiency in
Tfam mRNA could also be a sign of reduced mitochondrial number in vivo, as has been reported in dopaminergic neurons versus non-dopaminergic neurons of the mouse SNc [
235]. However, experiments with cultured SNc dopaminergic neurons actually suggest that SNc dopaminergic neurons have higher basal oxidative phosphorylation rates, higher mitochondrial density, and increased ROS production with respect to relatively resistant dopaminergic neurons derived from the ventral tegmental area [
236]. Further experiments are required to resolve these findings; if dopaminergic neurons do, in fact, have reduced mitochondrial density as compared to other neurons, pharmacological strategies which are meant to target existing mitochondria may not be very effective in preventing dopaminergic cell loss.
6.3. Developmental Disorders
The observations of PGC-1α enrichment in cortical PV-IN populations and its requirement for the expression of developmentally regulated PV-IN-enriched genes raise questions about its potential involvement in developmental disorders. In fact, the chromosomal location including
PPARGC1A has been linked to risk for schizophrenia and bipolar disorder [
237,
238,
239,
240], and a recent genome-wide association study found a significant association between schizophrenia and
rs215411, a single-nucleotide polymorphism ~370 kb downstream of
PPARGC1A [
241,
242].
A potential role for PV-IN dysfunction in schizophrenia pathophysiology was first suggested by a study indicating a reduction in
PVALB mRNA expression in cortical interneurons without a reduction in PV-IN number [
243]. Follow-up studies confirmed this reduction in
PVALB, accompanied by reductions in
LHX6 [
244], which encodes a transcription factor involved in PV-IN specification during development [
245]. Subsequent studies demonstrated that the deletion of the obligatory subunit of the NMDA receptor, NR1, from PV-INs in mice can generate the behavioral characteristics of autism and schizophrenia, including abnormalities in sensorimotor gating, social behavior, and working memory [
77,
246,
247,
248,
249,
250], with enhanced cortical and hippocampal gamma rhythms at baseline [
77,
250,
251]. Exposure of these mice to social isolation causes an age-related reduction in the expression of
Ppargc1a mRNA and PGC-1α protein in the cortex of these mice, associated with an increase in oxidative stress markers in PV-INs [
59].
With these studies in mind, we explored the expression of PGC-1α and PGC-1α-dependent genes in postmortem cortex of patients with schizophrenia [
252]. We found that while
PPARGC1A mRNA expression was not changed, mRNA for
PVALB,
SYT2,
NEFH, and
CPLX1, the PGC-1α-dependent transcripts identified in ([
70]), were reduced. Interestingly, bioinformatics analysis of the promoters of these genes revealed enrichment of consensus binding sites for NRF-1, and
NRF1 mRNA expression was reduced in the cortex of these patients [
252]. It’s possible that NRF-1 can mediate PGC-1α-dependent or independent regulation of neuron-enriched genes; while NRF-1 was initially identified as one of seven factors with consensus binding sites enriched in ubiquitously-expressed genes [
253], multiple studies have shown that NRF-1 and/or NRF-2 are involved in the regulation of genes enriched in neurons such as neuronal nitric oxide synthase [
254], AMPA receptor subunit 2 (
GRIA2; [
255,
256]), NMDA receptor subunits NR1 (
GRIN1) and NR2B (
GRIN2B; [
257]) and GABA receptor B1 (
GABRB1; [
258]). NRF-1 could serve as a critical transcription factor for the induction of cell-type-specific programs, depending on what combination of coactivators are present.
Along those lines, a disruption in transcriptional programs for PV-IN synaptic maturation and mitochondrial function has been found in postmortem tissue from patients with autism, bipolar disorder, and schizophrenia [
259,
260], with reduced abundance of mitochondrial electron transport chain protein complexes in both disorders [
261,
262]. A number of genes involved in mitochondrial function and responsive to PGC-1α overexpression in human neuroblastoma cells [
70] are reduced, including
ATP5A1 [
259,
260], glutamic-oxaloacetic transaminase 1 (
GOT1; [
259]),
IDH3A [
260], and malate dehydrogenase 1 (
MDH1; [
260]). These findings are concordant with reports of mitochondrial dysfunction in autism (reviewed in [
263] and [
264]) and schizophrenia (reviewed in [
265]). Transcriptional studies also reported decreased expression of
PVALB [
259,
260],
SYT2,
NEFH, and glutamic acid decarboxylase 1 and 2 (
GAD1 and
GAD2) [
259], indicating either the loss of expression of these markers and mitochondrial genes in existing interneuron populations or the loss of these neurons.
A limitation of these studies is that the majority of transcriptional profiling was performed in brain homogenates, making it challenging, if not impossible, to determine whether mitochondrial transcripts are downregulated in specific cell types. Recently, a human pluripotent cell line protocol was developed to generate homogeneous cultures of cells with the properties of PV or somatostatin-expressing cortical interneurons which originate from the medial ganglionic eminence [
266,
267]. Using this protocol, Ni et al. 2020 [
268] observed a reduction in the expression of genes involved in oxidative phosphorylation in iPSCs differentiated into interneurons but not pyramidal neurons from schizophrenia patients. These reductions were accompanied by decreased mitochondrial respiration and neurite arborization in interneuron cultures, which could be ameliorated by provision of a combination of alpha lipoic acid and acetyl-L-carnitine but not Coenzyme Q10 or N-acetyl cysteine. This study suggests that mitochondrial abnormalities in the cortex of patients with schizophrenia may arise from transcriptional alterations in interneurons; however, additional studies are needed to determine whether changes in transcription and function arise from disruption in nuclear or mitochondrial gene regulatory programs, as several of the reduced genes are expressed by the mitochondrial genome (ND2, ND5). Importantly, no statistically significant differences were noted in the percentage of neurons which express
GAD1, indicating that interneuron-like cells derived from schizophrenia patients were differentiating normally in culture. The authors do not report whether the expression of PV was different in the cultures, which is an important consideration when trying to compare iPSC studies with in vivo findings.
Mitochondrial dysfunction and oxidative stress has been reported for PV-INs in a number of models of neurodevelopmental disorders, including 22q11.2, 15q13.3, or 1q21 microdeletions, fragile X syndrome, and various developmental models of schizophrenia (reviewed in [
269]), suggesting that PV-IN impairment could be a common theme in neurodevelopmental disorders. Also, a recent article demonstrated that the deletion of COX10 from PV-expressing neurons in mice causes impairments in social behavior and sensori-motor gating, without affecting the total number of PV-INs in the cortex [
270]. Increased gamma oscillatory power is also observed, consistent with models of NR1 deletion in PV-INs (above). Altogether, these findings suggest that disruption of PGC-1α and/or mitochondrial genes in PV-INs can cause alterations in inhibitory neurotransmission in numerous developmental disorders, making it important to develop strategies to rescue these pathways.