The optic nerve: a "mito-window" on mitochondrial neurodegeneration.

Retinal ganglion cells (RGCs) project their long axons, composing the optic nerve, to the brain, transmitting the visual information gathered by the retina, ultimately leading to formed vision in the visual cortex. The RGC cellular system, representing the anterior part of the visual pathway, is vulnerable to mitochondrial dysfunction and optic atrophy is a very frequent feature of mitochondrial and neurodegenerative diseases. The start of the molecular era of mitochondrial medicine, the year 1988, was marked by the identification of a maternally inherited form of optic atrophy, Leber's hereditary optic neuropathy, as the first disease due to mitochondrial DNA point mutations. The field of mitochondrial medicine has expanded enormously over the last two decades and many neurodegenerative diseases are now known to have a primary mitochondrial etiology or mitochondrial dysfunction plays a relevant role in their pathogenic mechanism. Recent technical advancements in neuro-ophthalmology, such as optical coherence tomography, prompted a still ongoing systematic re-investigation of retinal and optic nerve involvement in neurodegenerative disorders. In addition to inherited optic neuropathies, such as Leber's hereditary optic neuropathy and dominant optic atrophy, and in addition to the syndromic mitochondrial encephalomyopathies or mitochondrial neurodegenerative disorders such as some spinocerebellar ataxias or familial spastic paraparesis and other disorders, we draw attention to the involvement of the optic nerve in classic age-related neurodegenerative disorders such as Parkinson and Alzheimer disease. We here provide an overview of optic nerve pathology in these different clinical settings, and we review the possible mechanisms involved in the pathogenesis of optic atrophy. This may be a model of general value for the field of neurodegeneration. This article is part of a Special Issue entitled 'Mitochondrial function and dysfunction in neurodegeneration'.

Introduction

In the year 1988, two neuro-muscular disorders were associated with the first pathogenic defects of mitochondrial DNA (mtDNA), starting the molecular era of mitochondrial medicine (Holt et al., 1988; Wallace et al., 1988). Leber's hereditary optic neuropathy (LHON), one of the two, is a form of optic nerve degeneration inherited through the maternal line associated with a point mutation affecting the MT-ND4 subunit gene of the respiratory complex I (Wallace et al., 1988). Since then, optic nerve atrophy has been recognized as a frequent hallmark of mitochondrial diseases and many neurodegenerative diseases have been discovered to have a primary molecular defect involving mitochondrial proteins (Babcock et al., 1997; Casari et al., 1998; Zuchner et al., 2004). Furthermore, mitochondrial dysfunction has been documented as a key pathogenic mechanism in many neurodegenerative disorders, even in those without a primary mitochondrial etiology (Schon and Area-Gomez, 2010; Vives-Bauza and Przedborski, 2011; Mochel and Haller, 2011). Finally, the introduction of new tools, such as optical coherence tomography (OCT) in ophthalmology, accurately assessed the optic nerve pathology in neurodegenerative diseases, such as Parkinson's disease, revealing patterns suggestive of mitochondrial neurodegeneration (La Morgia et al., 2012). Thus, the eye is a “mito-window” on the brain and this review will consider the retinal ganglion cells (RGCs) and related axons composing the optic nerve as a model system for mitochondrial neurodegeneration.

Optic neuropathy in mitochondrial diseases: clinical features

Non-syndromic: Leber's hereditary optic neuropathy (LHON) and dominant optic atrophy (DOA)

LHON and DOA are both characterized by an extreme selectivity of their tissue expression, which is limited to the RGCs and their axons in the optic nerve, with a preferential involvement of the small fibers in the papillomacular bundle that subserve central vision (Carelli et al., 2004; Yu-Wai-Man et al., 2011).

LHON patients, mostly young-adult males, typically undergo rapid and painless loss of central vision, which is bilateral in most cases, or involves sequentially one eye first and subsequently the other. The visual loss usually begins with defective color vision and central scotoma on the visual field. Loss of visual acuity stabilizes within the first year after onset, leaving the patient, in most cases, legally blind. Fundus changes characteristic of the subacute stage include peripapillary telangiectatic microangiopathy, swelling of the nerve fiber layer around the optic disk (pseudoedema) and lack of leakage on fluorescein angiography (in contrast to true edema). The optic disk appears hyperemic and eventually axonal loss in the papillomacular bundle leads to temporal pallor of the disk. In time, the optic disk turns completely pale.

The endpoint of LHON is usually optic atrophy with permanent loss of central vision and relative sparing of the pupillary light responses. Spontaneous recovery of visual acuity has been reported occasionally, with contraction of the scotoma or reappearance of small islands of vision within it (fenestration). The most favorable prognostic factors are young age of onset and type of pathogenic mutation, the 14484/MT-ND6 mutation being most commonly associated with spontaneous recovery (Carelli et al., 2011a). The chronic stage of LHON may be characterized by a further slowly progressing deterioration of visual functions, difficult to be objectively measured, but subjectively reported by patients and documented in a few cases studied post-mortem as evidence of still ongoing neurodegeneration (Carelli et al., 2002, 2004).

The recent use of OCT to measure retinal nerve fiber layer (RNFL) thickness provided anatomical quantitative features describing the subacute and chronic stages of LHON, as well as the unaffected mutation carriers and the recovery of vision. The crucial timing of LHON subacute conversion, detailed by OCT serial studies, demonstrated a precise pattern of axonal swelling followed by reduction, starting from the infero-temporal sector (papillo-macular bundle), followed by the superior and only lastly by the nasal quadrants, the latter usually the most spared by degeneration (Barboni et al., 2010a). This pattern matched that observed by the few postmortem studies of optic nerve specimens (Sadun et al., 2000; Carelli et al., 2002, 2004, 2009). In chronic LHON OCT showed a consistent decrease of RNFL thickness in all quadrants, which was significantly less severe in those patients recovering vision (Barboni et al., 2005). In unaffected mutation carriers the OCT showed a consistent increase of RNFL thickness (swelling) in the temporal/inferior quadrant, indicating a subclinical diseases of papillomacular bundle (Savini et al., 2005). Finally, the conformation of the optic nerve head has been substantiated by OCT as an “anatomical” risk factor for LHON, the small “at risk” disk being associated with conversion to affected status and a more severe final outcome (Ramos et al., 2009).

DOA patients, differently from LHON, present with slowly progressive, bilateral loss of central vision in childhood, accompanied by centrocaecal scotomas, impairment of color vision, temporal pallor of the optic disks and relative preservation of the pupillary reflex. The disease affects both sexes and is frequently so mild as to be recognized only during routine vision testing (school or driving license eye screenings). Its progression may vary within the same family, ranging from mild cases with visual acuity that stabilizes in adolescence, to slowly but relentlessly progressing cases, to severe cases with congenital optic atrophy. A distinct subgroup of LHON cases with childhood onset, especially if presenting with a slowly progressive clinical course, may pose serious problems of differential diagnosis with DOA (Barboni et al., 2006).

Despite the remarkably different clinical course, the endpoint of the pathological process in DOA is clinically indistinguishable from that in LHON, with a possible end-stage optic disk excavation, which may also be reported in LHON. The spontaneous recovery of visual acuity in DOA patients has not been reported, with a rare exception (Carelli et al., 2004; Yu-Wai-Man et al., 2011; Cohn et al., 2008; Cornille et al., 2008).

OCT studies in DOA defined quantitatively the optic atrophy, showing a decreased RNFL thickness, with a smaller average optic disk size, which suggests a reduced complement of axons at birth, compatible with severe congenital disease as a frust form of optic nerve hypoplasia (Barboni et al., 2010b; Barboni et al., 2011). The very few post-mortem studies of DOA showed selective loss of RGCs, particularly in the macular area, optic nerve axonal loss and demyelination, especially in the temporal aspect suggesting vulnerability of the papillomacular fibers as in LHON (Kjer et al., 1983).

An interesting feature shared by LHON and DOA is the occurrence, in a subset of cases, of white matter lesions, which may be indistinguishable from multiple sclerosis (MS) in terms of clinical and laboratory expression (Harding et al., 1992; Verny et al., 2008). This association of MS-like features with mitochondrial optic neuropathies seems to be causally related and not merely a coincidental occurrence of two overlapping diseases (Carelli and Bellan, 2008).

Syndromic: the overlapping LHON/MELAS/Leigh syndrome and DOA plus (OPAopathies)

LHON-like optic neuropathy has been described frequently in association with spastic dystonia and bilateral lesions in the basal ganglia, mainly bilateral striatal necrosis (Bruyn and Went, 1964; Novotny et al., 1986). Optic atrophy is also a frequent, but mostly overlooked feature of Leigh syndrome, a severe pediatric condition characterized by bilateral lesions of the central nervous system, usually extending from the basal ganglia to the brainstem (Cavanagh and Harding, 1994). Furthermore, LHON-like optic neuropathy is increasingly recognized in association with stroke-like episodes, typical for the MELAS syndrome (mitochondrial encephalomyopathy, lactic acidosis, stroke-like episodes) (Wallace, 1970; Pulkes et al., 1999). We recently reviewed this topic (Carelli et al., 2009) and emphasized that in many families the same mtDNA mutation, invariably affecting one of the ND subunits of complex I, may lead to each of the above mentioned clinical features alone or in combination in a bridging phenotype of LHON/MELAS/Leigh syndrome with the possible common pathogenic feature of mitochondrial vasculopathy (Pulkes et al., 1999; Chol et al., 2003; Blakely et al., 2005; Spruijt et al., 2007). Prominent optic atrophy in the contest of mitochondrial encephalomyopathy has also been associated with cytochrome c oxidase subunit I (COX I) mutation or with mtDNA deletion (Bruno et al., 1999; Rötig et al., 1993).

In 2008, we and others described the occurrence of a multisystemic disorder characterized by severe optic atrophy and sensorineural deafness, associated with cerebellar ataxia, axonal sensory–motor polyneuropathy, and late chronic progressive external ophthalmoplegia (CPEO) and ptosis due to a subgroup of missense mutations in the optic atrophy 1 (OPA1) gene, which was defined as DOA “plus” syndrome (Amati-Bonneau et al., 2008; Hudson et al., 2008). The most important observation in these patients was that they all had a mitochondrial myopathy with accumulation of multiple mtDNA deletions, thus implicating the OPA1 gene function in mtDNA maintenance. Interestingly, a single report recently presented a similar phenotype associated with compound heterozygote mutations in the polymerase gamma (POLG1) gene, demonstrating a genetic heterogeneity of optic atrophy associated with mitochondrial myopathy and mtDNA multiple deletions (Milone et al., 2011). Over recent years, other phenotypes have been described within the frame of the OPA1-related DOA “plus” syndrome with mtDNA multiple deletions, including MS-like features, Behr-like spastic paraparesis and cases with absent or subclinical ocular involvement, thus defining an increasingly large spectrum of “OPAopathies” (Yu-Wai-Man et al., 2010; Milone et al., 2009; Marelli et al., 2011; Pretegiani et al., 2011; Schaaf et al., 2011).

Remarkably, also mutations in mitofusin 2 (MFN2), another protein involved in mitochondrial fusion as OPA1 (see following sections), have been found in Charcot–Marie–Tooth axonal neuropathy type 2A (CMT2A) with optic atrophy, fulfilling the diagnostic criteria for Hereditary Motor Sensory Neuropathy type VI (HMSNVI) (Zuchner et al., 2004, 2006). These HMSNVI/CMT2A patients mostly presented with subacute visual loss, central scotoma, color vision defects and optic atrophy. They may experience, over years, variable degrees of recovery of visual acuity, sometimes reverting back to near normal visual function (Zuchner et al., 2006). Of note, a single family was recently reported with a MFN2 mutation segregating with optic atrophy beginning in early childhood, associated with axonal neuropathy and mitochondrial myopathy with mtDNA multiple deletions in adult life, closely resembling the DOA “plus” phenotype usually associated with OPA1 mutations, thus demonstrating genetic heterogeneity (Rouzier et al., 2012).

Intriguingly, optic atrophy may also complicate the clinical picture of families with autosomal recessive spastic paraplegia due to mutations in the spastic paraplegia 7 (SPG7) gene encoding for paraplegin, a mitochondrial metalloprotease of the AAA family, but unfortunately, description of the ophthalmological features in these patients is wanting (Casari et al., 1998).

Optic neuropathy is also a common feature in other multi-systemic mitochondrial disorders due to nuclear gene mutations, such as Friedreich ataxia (recessive-FRDA) (Campuzano et al., 1996; Babcock et al., 1997), the Behr-like Costeff syndrome (recessive-OPA3) (Anikster et al., 2001), recessive optic atrophy (OPA7/ROA1) with deafness (Hanein et al., 2009; Meyer et al., 2010), X-linked deafness–dystonia–optic atrophy syndrome (Mohr–Tranebjaerg syndrome) (Jin et al., 1996; Koehler et al., 1999) and a recessive syndrome of late-onset optic atrophy and ataxia with mutations in complex II (Taylor et al., 1996; Birch-Machin et al., 2000). This latter entity, observed in a single family to date, has an interesting overlap with FRDA, sharing both disorders complex II deficiency, which in FRDA is extended to all enzymes with sulfur–iron clusters. They also share a similar pattern of visual loss, starting from the periphery of the visual field, which is not canonical for mitochondrial disorders (Fortuna et al., 2009). Further, two interesting reports mention the occurrence of optic atrophy in very severe infantile encephalopathies due to novel mutations in two important mitochondrial proteins, the mitochondrial fission dynamin-1-like protein (DNML1) and the apoptosis induction factor (AIF) (Waterham et al., 2007; Ghezzi et al., 2010).

Last, we mention the recessive Wolfram syndrome, characterized by diabetes insipidus, diabetes mellitus, optic atrophy, and deafness (DIDMOAD), which has been controversially included as a mitochondrial disorder (Vora and Lilleyman, 1993; Bundey et al, 1993; Barrett et al., 2000). Causative mutations for Wolfram syndrome have been reported in two different genes, WFS1 and CISD2 (Inoue et al., 1998; Amr et al., 2007), encoding proteins localized respectively on endoplasmic reticulum (ER) and on mitochondrial outer membrane (OM) and ER (Chen et al., 2009). Given the prevalent mitochondrial localization of CISD2 and the possible role played by both WFS1 and CISD2 in calcium homeostasis, the debate on mitochondrial involvement in the pathogenic mechanism of Wolfram syndrome remains open (Kanki and Klionsky, 2009).

Optic neuropathy genetics

Mitochondrial inheritance

The maternal inheritance of LHON, clearly recognized a few decades ago, led to the identification of the frequent point mutations at positions 11778/MT-ND4, 3460/MT-ND1, and 14484/MT-ND6, which are found in over 90% of LHON cases worldwide (Carelli et al., 2004; Yu-Wai-Man et al., 2011). LHON can be also due to rare pathogenic mtDNA point mutations affecting different subunits of complex I, more commonly the MT-ND6 and MT-ND1 subunit genes (Chinnery et al., 2001; Valentino et al., 2004) (see Table 1). The mtDNA pathogenic mutations, which are in most cases homoplasmic (100% of the mtDNA molecules are mutated), do not explain at least two features of LHON: the male prevalence and the incomplete penetrance. Thus, the mtDNA mutation is a necessary but not a sufficient condition for the pathology, and additional genetic determinants, such as nuclear modifying genes, have been postulated (Carelli et al., 2003). The X chromosome has been considered for a long time as a good candidate to explain both gender difference and variable penetrance by hypothesizing a two-loci model (mitochondrial pathogenic plus nuclear modifying) (Bu and Rotter, 1991). However, despite the identification of some loci, no modifying genes on the X-chromosome have been truly established to date (Hudson et al., 2005; Shankar et al., 2008; Ji et al., 2010).

Table 1

Pathogenic mtDNA mutations associated with Leber hereditary optic neuropathy, all reported by MITOMAP, 2011, except a reported by Zhang et al., 2012 and b by Zhao et al., 2009.

	Mutation	Gene	Change
Common variants (~ 90%)	11778 G > A	MT-ND4	R340H
	3460 G > A	MT-ND1	A52T
	14484 T > C	MT-ND6	M64V
Rare variants (~ 10%)	3635 G > A	MT-ND1	S110N
	3733 G > A		E143H
	4171 C > A		L289M
	4640 C > A	MT-ND2	I57M
	5244 G > A		G259S
	10663 T > C	MT-ND4L	V65A
	10680 G > Aa		A71T
	14482 C > A, C > G	MT-ND6	M64I
	14495 A > G		L60S
	14502 T > Cb		I58V
	14568 C > T		G36S
Reported rare variants in ND subunits	3700 G > A	MT-ND1	A112T
	3736 G > A		V144I
	4025 T > C		T240M
	4160 T > C		L285P
	4640 C > A	MT-ND2	I57M
	4852 T > A		L128Q
	5244 G > A		G259S
	10237 T > C	MT-ND3	I60T
	10543 A > G	MT-ND4L	H25R
	10591 T > G		F41C
	11253 T > C	MT-ND4	I165T
	11696 G > A		V312I
	11874 C > A		T372N
	12782 T > G	MT-ND5	I149S
	12811 T > C		Y159H
	12848 C > T		A171V
	13051 G > A		G239S
	13379 A > C		H348P
	13637 A > G		Q434R
	13730 G > A		G465E
	14279 G > A	MT-ND6	S132L
	14325 T > C		N117D
	14498 C > T		Y59C
	14596 A > T		I26M
Reported rare variants in other mtDNA genes	7623 C > T	MT-CO2	T13I
	8668 T > C	MT-ATP6	W48R
	8836 A > G		M104V
	9016 A > G		I164V
	9101 T > C		I192T
	9660 A > C	MT-CO3	M152L
	9738 G > T		A178S
	9804 G > A		A200T
	14831 G > A	MT-CYB	A29T
	15674 T > C		S310P

The only strong evidence of a genetic modifier is currently the association of LHON with a European mtDNA background, known as haplogroup J, which increases the penetrance of the 11778/MT-ND4 and 14484/MT-ND6 LHON mutations (Carelli et al., 2006; Hudson et al., 2008). The modifying effect of haplogroup J is thought to be due to specific arrays of complex I and III non-synonymous polymorphisms characterizing sub-haplogroup J1 for the 14484/ND6 mutation and sublineages J1c and J2b for the 11778/MT-ND4 mutation (Carelli et al., 2006). A recent study, investigating OXPHOS efficiency in cybrid cells carrying different mtDNA haplogroups from normal people, revealed subtile defects in the OXPHOS capacity of cells carrying the haplogroup J mtDNA, which leads support to the modifying effect of this mitochondrial background on LHON penetrance (Gómez-Durán et al., 2012).

In addition to the primary LHON mutations, many other point mutations affecting ND subunits of complex I have been reported in association with overlapping phenotypes, such as the LHON/MELAS/Leigh syndrome. These mutations affect most frequently the MT-ND5, MT-ND6 and MT-ND1 subunit genes, but also MT-ND4, MT-ND3 and MT-ND2 (see Table 2). Some cases of syndromic and non-syndromic optic atrophy are associated with mtDNA mutations other than the ND subunits genes of complex I, such as, for example, in cytochrome b, COX I and ATPase 6 genes (see Table 1).

Table 2

Pathogenic mtDNA mutations associated with overlapping phenotypes.

Mutation	Gene	Clinical phenotype	Reference
3376 G > A	MT-ND1	LHON/MELAS	Blakely et al. (2005)
3697 G > A		LHON/MELAS	Spruijt et al. (2007)
10197 G > A	MT-ND3	LHON/dystonia	Wang et al. (2009)
13042 G > A	MT-ND5	LHON/MELAS/MERFF	Valentino et al. (2006)
13045 A > C		MELAS/LHON/Leigh	Liolitsa et al. (2003)
13513 G > A		LHON/MELAS/Leigh	Blok et al. (2007)
13514 G > A		LHON/MELAS	Corona et al. (2001)
14459 G > A	MT-ND6	LHON/dystonia	Jun et al. (1994)
14600 G > A		LHON/MELAS/Leigh	Malfatti et al. (2007)

Mendelian inheritance

A growing list of optic atrophy loci (OPA) associated with Mendelian transmission of the trait (recessive, dominant and X-linked) has been building in recent years, with only a few mutant genes already identified (see Table 3). In addition to those with a certified or highly probable mitochondrial etiology, optic atrophy is a frequent feature in many neurodegenerative monogenic inherited disorders, for which the link to mitochondria remains elusive (see Table 4). However, the pattern of preferential involvement of the papillomacular bundle, a hallmark of mitochondrial etiology, may anticipate that many of these disorders will turn out to have a primary or secondary involvement of mitochondrial function (Winkelmann et al, 2012; Stricker et al, 2011).

Table 3

List of optic atrophy loci (OPA) associated with optic atrophy (OA).

Gene/locus	Protein	Location	Heritance	Phenotype	OMIM	Reference
OPA1	OPA1	3q29	Dominant	OA	#165500	Alexander et al. (2000); Delettre et al. (2000)
OPA2	–	Xp11.4–p11.21	X-linked	OA with early onset	#311050	Assink et al. (1997)
OPA3	OPA3	19q13.32	Dominant	OA with cataract	#165300	Reynier et al. (2004)
OPA4	–	18q12.2–q12.3	Dominant	OA	#605293	Kerrison et al. (1999)
OPA5	–	22q12.1–q13.1	Dominant	OA	#610708	Barbet et al. (2005)
OPA6	–	8q21–q22	Dominant	OA	#258500	Barbet et al. (2006)
OPA7/ROA1	TMEM126A	11q14.1	Recessive	OA	#612989	Hanein et al. (2009)
OPA8	–	16q21–q22	Dominant	OA with deafness	–	Carelli et al. (2011b)
WFS1	Wolframin	4p16.1	Dominant	OA with deafness	–	Rendtorff et al. (2011)

Table 4

List of major genes associated with optic atrophy syndromes and encoding for proteins with a putative or demonstrated mitochondrial function.

Gene	Protein	Location	Heritance	Phenotype	OMIM	Reference
OPA1	OPA1	3q29	Dominant	OPA1 plus	#125250	Amati-Bonneau et al. (2008)
OPA3	OPA3	19q13.32	Recessive	Costeff syndrome	#258501	Anikster et al. (2001)
FXN	Frataxin	9q21.11	Recessive	Fredreich ataxia	#229300	Fortuna et al. (2009)
TIMM8A	TIMM8A	Xq22	X-linked	Mohr–Tranebjærg syndrome	#304700	Tranebjaerg et al. (2001)
SPG7	Paraplegin	16q24.3	Recessive	Spastic paraplegia-7	#607259	Casari et al. (1998)
MFN2	Mitofusin 2	1p36.22	Dominant	Charcot–Marie–Tooth disease, type 2A2	#609260	Zuchner et al. (2006)
DNM1L	DNM1L	12p11.21	Dominant	Encephalopathy	#614388	Waterham et al. (2007)
WFS1	Wolframin	4p16.1	Recessive	Wolfram syndrome 1	#222300	Strom et al. (1998)
CISD2	CISD2	4q24	Recessive	Wolfram syndrome 2	#604928	Amr et al. (2007)
SDHA	Complex II	5p15	Dominant	Late-onset optic atrophy and ataxia	–	Birch-Machin et al. (2000)

Optic neuropathy in neurodegenerative diseases

The retinal and optic nerve involvement in major neurodegenerative disorders, such as Parkinson and Alzheimer disease, has been described decades ago (Bodis-Wollner and Yahr, 1978; Hinton et al., 1986). However, recent technical improvements in eye investigation, such as OCT, gave great impetus to re-investigation of retinal and optic nerve involvement in neurodegenerative diseases, the eye being the most accessible part of the central nervous system.

Parkinson disease (PD)

There have been several lines of evidence supporting the involvement of the retina and the optic nerve, as non-motor component, in the neurodegenerative process of PD (Bodis-Wollner, 2009; Archibald et al., 2009). In particular, impaired visual acuity (Jones et al., 1992), abnormal contrast sensitivity (Bodis-Wollner et al., 1987), color vision abnormalities (Silva et al., 2005) and optic nerve degeneration have been described in PD (Archibald et al., 2009).

Since 2004, OCT measurements in PD patients documented a reduction of the RNFL thickness in both early and more advanced stages of the disease. Some authors reported a preferential loss of fibers in the infero-temporal quadrants, consistent with the involvement of the papillo-macular bundle (Inzelberg et al., 2004; Yavas et al., 2007; Moschos et al., 2011). Macular and foveal abnormalities have also been documented (Hajee et al., 2009).

Our own study of 43 PD patients documented a subclinical optic neuropathy, affecting significantly only the nerve fibers entering the temporal quadrant of the optic disk (La Morgia et al., in press). The contralateral eye was more affected than the ipsilateral to the most affected body side, supporting an asymmetry of the neurodegenerative process that involves both the substantia nigra and the eye on the same side. These results identified in PD patients a pattern of axonal loss typically seen in LHON and DOA, where the papillo-macular bundle is characteristically susceptible (Carelli et al., 2004). Both diseases are associated with a complex I defect (Carelli et al., 1997, 1999; Zanna et al., 2008; Van Bergen et al., 2009; Carelli et al., 2007, 2009), which is also recognized as a key feature in the pathogenesis of PD, both in the sporadic cases as well as in the genetic forms (Schapira, 2008), linking the current results with the pattern observed in LHON and DOA.

Alzheimer disease (AD)

Complex visual complaints, including difficulties in reading or finding objects, depth perception, perceiving structure from motion, color recognition and contrast sensitivity abnormalities are frequent in AD (Guo et al., 2010; Sadun et al., 1987). In 1986 it was shown by Hinton and colleagues, in post-mortem retina and optic nerves of AD patients, that there was loss of RGCs and related axons, involving preferentially the M-cell system, in the apparent absence of neurofibrillary or amyloid angiopathy (Hinton et al., 1986; Sadun and Bassi, 1990). After this first report, other authors showed the presence of optic nerve degeneration, being more evident in the superior and inferior quadrants and in the foveal region in post-mortem tissues of AD patients (Blanks et al., 1996a, 1996b).

More recently, the use of OCT allowed not only the objective measurements of RNFL, but also of macular thickness for in-vivo quantitative evaluation of RGCs loss in AD. Studies with this approach reported a reduction in the average measurements (Parisi et al., 2001; Iseri et al., 2006; Paquet et al., 2007; Lu et al., 2010; Kesler et al., 2011), as well as in the superior quadrant (Parisi et al., 2001; Iseri et al., 2006; Berisha et al., 2007; Lu et al., 2010; Kesler et al., 2011) and in the nasal quadrant (Parisi et al., 2001; Iseri et al., 2006). Remarkably, these studies revealed the lack of involvement of the temporal quadrant. Overall, the subclinical optic neuropathy in AD seems to affect the magnocellular component of RGCs, as documented both histologically (Hinton et al., 1986; Sadun and Bassi, 1990) and more recently electro-physiologically (Sartucci et al., 2010). In this regard, it is of note that this pattern resembles that of glaucomatous optic neuropathy and, in fact, there is a documented evidence of increased prevalence of glaucoma in AD patients (Bayer et al., 2002).

Mitochondrial function and dysfunction in optic nerve neurodegeneration

The pathways leading to neurodegeneration of the RGCs and related optic nerve have been delineated over the last 20 years and include biochemical impairment of complex I, the consequent increase in reactive oxygen species (ROS) and activation of apoptosis. More recently, attention has been drawn to the role played by the mitochondrial network dynamics and mitophagy, as well as to the regulation of mitochondrial biogenesis as a compensatory strategy to mitochondrial dysfunction (Fig. 1).

Fig. 1

Schematic illustration of the genes mutated (blue) in mitochondrial optic neuropathies and their functions within mitochondria. Question marks indicate unknown localization and/or mitochondrial function.

Complex I

Mitochondria are the primary source of ATP providing the energy required for cell functioning through oxidative phosphorylation (OXPHOS). All the five large enzymatic complexes constituting the OXPHOS system are situated in the inner mitochondrial membrane (IM), where, together with two mobile electron carriers, ubiquinone (CoQ) and cytochrome c (cyt c), they couple electron transport to the generation of an electrochemical proton gradient across the IM. This gradient can drive the synthesis of ATP from ADP and inorganic phosphate at the level of Complex V (F0/F1-ATP-synthase), and, in addition, is used for various energy-dependent processes, such as nuclear protein import, Ca2 + uptake, and ion translocation (Koopman et al., 2010).

With approximately 45 subunits, Complex I (CI, NADH:ubiquinone oxidoreductase) is the largest of the respiratory multi-protein complexes; the subunits are encoded both by nuclear (38 subunits) and mitochondrial (7 subunits) genomes (Carroll et al., 2006), 14 subunits constituting the highly conserved catalytic core (Brandt, 2006). CI transfers two electrons from NADH to CoQ, generating ubiquinol (CoQH2) and producing the translocation of four protons from matrix to the inner membrane space (IMS) (Galkin et al., 1999). During this step, electrons may prematurely escape from the electron transport chain and react with oxygen to form superoxide (O2•−), making the CI a major site of production of this oxygen radical (Duchen, 2004).

Benard et al. (2006) observed different expression levels of CI, evaluated by quantification of the nuclear subunit NDUFA9 (NADH dehydrogenase ubiquinone 1 alpha subunit 9), in rat heart, muscle, brain, liver and kidney. They also found a strong correlation between CI expression and OXPHOS activity in these tissues: heart and muscle had the highest CI expression and OXPHOS activity, liver and kidney the lowest, and brain presented as intermediate CI expression and OXPHOS capacity (Benard et al., 2006). As a consequence, different tissues may show a different sensitivity to OXPHOS defects.

Dysfunction of CI is the most frequent defect of mitochondrial energetic metabolism (Smeitink et al., 2001; Distelmaier et al., 2009) and represents a common feature of mitochondrial optic neuropathies. LHON is due to three common mutations affecting mitochondrial CI subunits, which induce moderate to severe impairment of CI activity, depending on the mutation (Carelli et al., 2004). Studies with different CI inhibitors also indicated that all three mutations affect the interaction between CI and ubiquinone (Carelli et al., 2004) and consistently decrease ATP synthesis driven by complex I substrates, confirming that CI is defective in LHON (Baracca et al., 2005). Similarly, reduced ATP synthesis driven by CI substrates was found in patient-derived fibroblasts carrying different OPA1 mutations (Zanna et al., 2008; Chevrollier et al., 2008). Co-immunoprecipitation experiments showed that OPA1 directly interacts with CI subunits, possibly regulating its assembly and stabilization (Zanna et al., 2008). The existence of a common biochemical OXPHOS defect, shared by LHON and DOA, was further corroborated by in vivo 31P-MRS studies of skeletal muscle (Lodi et al., 1997, 2011).

The central theme of CI dysfunction in optic neuropathies also applies to the syndromic cases. For example, Leigh syndrome and the clinical overlapping phenotype of LHON/Dystonia/MELAS/Leigh syndrome have been associated with mutations in ND subunit genes, mainly MT-ND5, MT-ND6 and MT-ND1 (Carelli et al., 2009 and Table 2). In FRDA the reduced amounts of frataxin may lead to impaired CI, confirming its central role in the mitochondrial Fe–S cluster biosynthesis (Stemmler et al., 2010). Also in MTS, Mohr Tranebjaerg syndrome, an altered import of nuclear-encoded subunits of CI, may lead to the loss of RGCs and depletion of axons in the optic nerve described in this syndrome (Tranebjaerg et al., 2001). Finally, defective paraplegin was demonstrated to induce reduction of CI activity and increased sensitivity to oxidative stress (Casari et al., 1998), as well as a CI dysfunction may be part of the pathogenesis in PD, explaining the mitochondrial pattern of axonal loss in the optic nerve (La Morgia et al., in press).

However, CI deficiency is not the only exclusive biochemical OXPHOS defect that leads to mitochondrial optic neuropathy. Other mtDNA mutations in structural genes such as cytochrome b, COX I, ATPase 6 (Table 1) or mutations in tRNA genes have been associated with syndromic and non-syndromic optic atrophy (Chinnery et al., 1997). Furthermore, complex II deficiency, associated with a genetic defect in the nuclear-encoded flavoprotein subunit of complex II, may be associated with optic atrophy (Taylor et al., 1996; Birch-Machin et al., 2000). Of note, besides CI, biochemical investigations of OPA1 mutant cells from DOA patients also showed a defective COX activity (Chevrollier et al., 2008; Valerio Carelli and Anna Ghelli, unpublished data).

Reactive oxygen species

It is well known that mitochondria produce reactive oxygen species (ROS) as a consequence of electron leak during their transport. Superoxide radicals represent the primary ROS produced and complex I and complex III are the major superoxide (O2•−)-producing sites in mitochondria (Lenaz, 1998). The O2•− is rapidly converted into hydrogen peroxide (H2O2) by manganese superoxide dismutase (MnSOD) and is further metabolized by glutathione peroxidase (GPX) to H2O. In the presence of transition metals, the H2O2 can also form the hydroxyl radical (OH•) through the Fenton reaction. Moreover, O2•− may react with nitric oxide (NO•) leading to peroxynitrite (ONOO•) production.

ROS play a crucial role in regulating several cellular processes. ROS, when produced at low physiological concentration, are important messengers in cell signaling (Lenaz, 2012). However, if ROS levels overcome a threshold they can disrupt normal cellular function damaging DNA, proteins and lipids (Handy and Loscalzo, 2012). Oxidative damage to mtDNA may cause an increased rate of mitochondrial genomic instability and accumulation of somatic mutations, ultimately leading to respiratory dysfunction and possibly contributing to cancer, premature aging and neurodegenerative diseases (Bohr, 2002). Excessive ROS production may also cause local damage to the Fe–S clusters of respiratory enzymes (complexes I, II and III), as well as to tricarboxylic acid cycle enzymes (aconitase). Peroxynitrite can nitrate tyrosine residues or thiolic groups of nearby proteins and both CI and MnSOD have been reported to be damaged by this process (Rötig et al., 1997; Melov et al., 1999). Another important damaging process is lipid peroxidation, which may affect several mitochondrial functions including OXPHOS, inner membrane barrier properties, maintenance of mitochondrial membrane potential (Δψm), and mitochondrial Ca2 + buffering capacity (Zhang et al., 1990; Albano et al., 1991; Reed, 2011). In particular, an important target of ROS is cardiolipin (CL), an IM phospholipid involved in mitochondrial-dependent apoptosis and mitochondrial stability and dynamics (Perier et al., 2005; Ban et al., 2010; Paradies et al., 2010), and its peroxidation plays a critical role in several patho-physiological mechanisms (Paradies et al., 2010). ROS may also act as signaling molecules regulating function or expression of many proteins, including transcription factors (p53, JAK, NF-kb, HIF-1, etc.), kinases (p38, PKA, PKC, ERK, JNK, etc.), phosphatases (i.e. PTEN), ion channels and other mitochondrial proteins involved in apoptosis, dynamics, mitochondrial biogenesis, etc. (Koopman et al., 2010). Thus, the balance between ROS production and their removal by antioxidant enzymes (superoxide dismutases, glutathione peroxidases and peroxiredoxins) is critical (Handy and Loscalzo, 2012).

Increase of ROS production, as a consequence of CI dysfunction, has been observed in osteosarcoma (143B TK-)-derived (Beretta et al., 2004) and in neuronal (NT2)-derived (Wong et al., 2002) cybrids carrying LHON mutations. Increased oxidative stress in DOA has recently been shown in a Drosophila model carrying a mutation in the Opa1 gene and a multi-system phenotype (Yarosh et al., 2008; Tang et al., 2009). Oxidative stress could be decreased by treating these flies with vitamin E or superoxide dismutase 1 (SOD1), as well as by over-expressing human SOD1, partially reversing the pathological phenotype, and providing a strong indication of the possible role played by ROS in DOA (Yarosh et al., 2008; Tang et al., 2009). This issue has not been properly addressed in mammals yet, but appropriate experimental investigations of DOA in rodents model and human patients are warranted. Oxidative stress plays an important role also in FRDA, where the excess of free iron, due to defective frataxin, leads to an increased production of H2O2 (Armstrong et al., 2010). The ROS damaging role in FRDA has been also confirmed in a Drosophila model, in which H2O2 scavengers provided protective effects on the phenotype, contrary to O2•− scavengers (Anderson et al., 2008).

Overall, increased oxidative stress is a common feature in the pathophysiology of mitochondrial optic neuropathies.

Mitochondrial dynamics

Mitochondria are highly dynamic organelles able to fuse and divide, depending on the energy requirement of the cell or in response to endogenous and exogenous stimuli. The morphology of the mitochondrial network results from a balance between the two antagonistic mechanisms of fusion, allowing an extended and interconnected reticular network, and fission that generates many mitochondria as distinct organelles (Westermann, 2010). While the large network resulting from fusion metabolically supports the cells and regulates the response to calcium signals (Skulachev, 2001; Szabadkai et al., 2006), fission is essential for proliferation and distribution of organelles in daughter cells after mitosis and is required for the elimination of old or damaged mitochondria through the autophagic process known as mitophagy (Palmer et al., 2011). Furthermore, fission is also required for axoplasmic transport of mitochondria in neurons (Saxton and Hollenbeck, 2012).

Many proteins regulating mitochondrial fusion and fission have been identified in the last 15 years, but the scenario is continuously evolving. Mitochondrial fusion is made possible by the coordinated action of pro-fusion OM (mitofusins, MFN1 and MFN2) and IM (OPA1) proteins, and additional scaffolding (Prohibitin 2, stomatin like protein-2) proteins in the IM. MFN1, MFN2 and OPA1 are all transmembrane proteins with a GTPase activity (Rojo et al., 2002; Santel and Fuller, 2001; Satoh et al., 2003) and both MFN2 and OPA1 show additional non-fusion functions. In fact, MFN2 is involved in mitochondrial biogenesis (Cartoni et al., 2005; Zorzano et al., 2010) and in tethering mitochondria to the endoplasmic reticulum (de Brito and Scorrano, 2008), whereas OPA1 participates to the cristae structure maintenance (Olichon et al., 2003; Griparic et al., 2004) and remodeling to prevent apoptosis (Frezza et al., 2006; Cipolat et al., 2006) and ultimately is involved in mtDNA maintenance (Elachouri et al., 2011).

OPA1 exists in 8 different isoforms, generated by the alternative splicing of exon 4, exon 5 and exon 5b, that can be present or absent (Olichon et al., 2007a). The pattern of expression of these transcripts has been demonstrated to be tissue-specific in humans, with isoform 1 (containing only exon 4), isoform 3 (containing only exon 4b) and isoform 5 (containing exons 4 and 4b) being predominantly expressed in brain (Olichon et al., 2007a). Thus, the understanding of the physiological function of each OPA1 isoform and the characterization of the expression levels of each variant in human RGCs might help to clarify the pathogenetic mechanism responsible for OPA1-associated DOA and to work out innovative therapeutic strategies, such as replacement gene therapy.

Furthermore, OPA1 isoforms undergo proteolytic cleavage producing long isoforms (l-OPA1), anchored to the IM, and soluble short isoforms (s-OPA1), (Satoh et al., 2003; Ishihara et al., 2006; Olichon et al., 2002). Several proteases regulating mammalian OPA1 processing have been identified (PARL, m-AAA protease, OMA1, Yme1) and the balance of OPA1 processed isoforms has been demonstrated to affect the mitochondrial morphology, acting as an important regulatory mechanism of mitochondrial fusion (Cipolat et al., 2006; Griparic et al., 2007; Ishihara et al., 2006; Song et al., 2007; Ehses et al., 2009; Head et al., 2009).

In addition to proteases involved in the maintenance of OPA1 isoforms' balance, other proteins that regulate mitochondrial fusion have been identified. Most of these proteins are regulators of other processes, such as apoptosis (Bax, Bak), mitophagy (Parkin), and remodeling of mitochondrial membrane lipids (MitoPLD, Lipin 1b), whereas mitofusin binding protein (MIB) negatively regulates MFN1 but is not involved in other mitochondrial processes (Palmer et al., 2011).

Different proteins have been proposed to contribute to mitochondrial fission regulation, but so far only DNM1L has been demonstrated to play a direct role in fission (Mozdy et al., 2000; Smirnova et al., 1998, 2001). DNM1L is a cytosolic protein with GTPase activity, recruited on mitochondrial OM when fission occurs. DNM1L promotes OM fission through the assembly of multimeric ring complexes, coupling GTP hydrolysis with mitochondrial membrane constriction and fission (Zhang and Hinshaw, 2001; Smirnova et al., 2001). Some proteins have been proposed to mediate the recruitment of DNM1L on the OM, including mitochondrial fission protein 1 (FIS1) (Zhang and Chan, 2007), mitochondrial fission factor (MFF) (Otera et al., 2010; Gandre-Babbe and van der Bliek, 2008) and MiD proteins (Palmer et al., 2011), although recently the interaction with FIS1 and its role in mitochondrial fission in mammals have been debated (Palmer et al., 2011). Little is known about IM division, however MTP18 (mitochondrial fission protein 1), a phosphatidylinositol 3-kinase-dependent protein in IM membrane, has been proposed to play a role in this process (Tondera et al., 2004, 2005).

What is known about the regulation of mitochondrial fission is so far limited to DNM1L post-translational modification, such as phosphorylation, S-nitrosylation, ubiquitanation and sumoylation (Palmer et al., 2011).

The distribution of mitochondria in neuronal cells has been demonstrated to have an important role in regulating the synaptic activity. Abnormal mitochondrial dynamics and distribution have been observed in neurodegenerative disorders like Alzheimer, Parkinson, Huntington and Familial Amyotrophic Lateral Sclerosis (FALS) (Reddy et al., 2012; Büeler, 2009; Burbulla et al., 2010; Shirendeb et al., 2011; Magrané et al., 2012). This becomes crucial in the context of OPA1-related DOA, where RGCs are primed to apoptotic cell death, which leads to optic nerve degeneration (Carelli et al., 2009).

DOA fibroblasts cultured in glucose medium may present a mitochondrial network ranging from normal to highly fragmented depending on the OPA1 mutation (Chevrollier et al., 2008; Olichon et al., 2007b). Balloon-like structures and inhibition of mitochondrial fusion were observed in fibroblasts carrying OPA1 nonsense mutations cultured in galactose medium (Zanna et al., 2008). Mitochondrial structural alterations have been frequently reported in myotubes (Spinazzi et al., 2008) and skeletal muscle from DOA patients (Amati-Bonneau et al., 2008), as well as in OPA1 mouse models (Alavi et al., 2007; Davies et al., 2007). One specific OPA1 mutation showed altered fusion activity without affecting bioenergetics or increasing sensitivity to apoptosis, suggesting that defective pro-fusion activity is the only pathogenic aspect (Spinazzi et al., 2008).

Visual loss is also a feature of CMT2A neuropathy, emphasizing the link between optic neuropathies and dysfunction of mitochondrial dynamics (Carelli et al., 2009). However, studies on MFN2-mutant CMT2A fibroblasts failed to show defective mitochondrial fusion, even if a bioenergetic defect was observed (Loiseau et al, 2007; Amiott et al., 2008). Complementation by the fellow MFN1 protein possibly rescued the phenotype in these cells, as demonstrated in single and double mitofusin knockout mouse embryonic fibroblasts (MEFs) (Detmer and Chan, 2007). Thus, we can speculate that in CMT2A patients, RGCs might be more sensitive to MFN2 dysfunction and suffer abnormal mitochondrial fusion, due to low MFN1 expression levels.

A dominant negative heterozygous DNM1L mutation, recently found in a patient affected by severe infantile encephalopathy and optic atrophy, showed a highly filamentous interconnected mitochondrial network in the patient-derived fibroblasts (Waterham et al., 2007). This case represents a demonstration that increased mitochondrial fusion activity may also have dramatic effects, suggesting that the imbalance of mitochondrial fusion–fission plays a central role in the pathogenesis of mitochondrial optic neuropathies.

Finally, a peculiar form of DOA with cataract has been associated with heterozygous mutations in the OPA3 gene (Reynier et al., 2004; Verny et al., 2005), whereas recessive mutations are responsible for Costeff syndrome, a 3-methilglutaconic aciduria with a Behr-like phenotype that includes optic atrophy (Anikster et al., 2001; Kleta et al., 2002). The function of this mitochondrial protein remains poorly understood, but a putative role in fusion–fission of mitochondria remains a strong possibility (Ryu et al., 2010).

Mitochondrial apoptosis

Apoptosis is a form of programmed cell death essential for homeostasis, which is frequently dysregulated in human pathologies such as cancer, neurodegenerative diseases or viral infections (Meier et al., 2000; Vaux and Korsmeyer, 1999). Mitochondria are involved in the so-called intrinsic pathway of apoptosis, executing the release of soluble proteins, including cyt c, from the intermembrane space to switch on the caspase activation in the cytosol. The release of these proteins occurs after the loss of mitochondrial OM integrity (mitochondrial outer membrane permeabilization, MOMP), which, in vertebrates, is regulated by the proapoptotic Bcl-2 family members (Martinou and Youle, 2011).

Both external and internal stimuli can activate different signaling pathways, transducing the signal to mitochondria by Bcl-2 proteins (Jourdain and Martinou, 2009). The mechanism by which the Bcl-2 proteins induce OM permeabilization and release the apoptogenic factors remains unclear. The current hypothesis postulates the opening of one of two different channels: the permeability transition pore (PTP) in the IM and the mitochondrial apoptosis-induced channel (MAC) in the OM (Galluzzi and Kroemer, 2007; Kinnally and Antonsson, 2007). Whatever is the mechanism of mitochondrial permeabilization, it implies the release of apoptogenic factors (cyt c, AIF, endonuclease G, Smac/DIABLO and Omi/HtrA2) (Wang and Youle, 2009).

The last step in apoptosis is the activation of specific proteases called caspases (cysteine aspartyl-specific proteases) that cleave their substrates at aspartic acid (Asp) residues (Thornberry and Lazebnik, 1998; Cryns and Yuan, 1999). These intracellular proteases are activated by proteolytic cleavage at conserved Asp residues and induce proteolytic cascades, where they activate themselves and each other, finally cleaving many substrates, promoting degradation of cell cytoskeleton and nuclear envelope, loss of adhesion, mitochondrial, Golgi and endoplasmic reticulum fragmentation and DNA condensation and fragmentation, ultimately leading to cell death (Taylor et al., 2008).

Apoptosis is believed to be the final stage of the neurodegenerative processes affecting RGCs in mitochondrial optic neuropathies. However, direct evidence of this is poor (Cordeiro et al., 2010). Studies on LHON cybrids, using Fas-induced apoptosis, showed that LHON mutant cells were more prone to undergo apoptotic cell death than control cells (Danielson et al., 2002). Similarly, LHON cybrids grown in galactose medium, which forces oxidative metabolism, displayed massive, apoptotic cell death, whereas control cybrids were still able to grow at a reduced rate (Ghelli et al., 2003). Under this paradigm, the apoptotic death in LHON cybrids was demonstrated to be caspase-independent, mediated by AIF and Endo G (Zanna et al., 2005). An increased sensitivity to staurosporine, a pro-apoptotic compound, was also found in fibroblasts from OPA1-mutant DOA patients (Olichon et al., 2007b), compatible with OPA1 regulation of cyt c release and apoptosis (Olichon et al., 2003, 2007a). A further link between OPA1 and apoptosis is suggested by its coimmunoprecipitation with AIF (Zanna et al., 2008). Furthermore, overexpression of the p.G93S/OPA3 mutant protein in HeLa cells led to spontaneous apoptosis (Ryu et al., 2010), confirming the previous observation of an increased sensitivity to staurosporine in fibroblasts carrying the p.G93S/OPA3 mutation (Reynier et al., 2004).

Mitochondrial biogenesis

Mitochondrial biogenesis is a complex process involving the coordinated expression of both nuclear and mitochondrial genes. Since the protein coding capacity of mtDNA is restricted to the expression of only 13 respiratory subunits, nuclear genes are extensively needed for oxidative phosphorylation, heme biosynthesis, mitochondrial protein import, and mtDNA transcription and replication. Expression of the mitochondrial proteome and control of mitochondrial biogenesis are highly regulated by transcription factors and transcriptional coactivators (Diaz and Moraes, 2008; Scarpulla, 2006).

The most important transcription factors involved in the mitochondrial–nucleus communication are the nuclear respiratory factors 1 and 2 (NRF-1, NRF-2) and the estrogen-related receptor (ERRα), whose expression is driven by the transcriptional coactivators belonging to the peroxisome proliferator-activated receptor γ-coactivator 1 (PGC-1) family (Scarpulla, 2002).

The PGC-1 family is composed of three members (PGC-1α, PGC-1β, PRC), which share a sequence homology and regulate several metabolic pathways such as cellular respiration, thermogenesis and hepatic glucose metabolism (Scarpulla, 2006; Kelly and Scarpulla, 2004). Although all these factors can stimulate mitochondrial biogenesis, PGC-1α is mainly involved in the regulation of gluconeogenesis and PGC-1β in the regulation of β-oxidation of fatty acids, and PRC in the coordination of nuclear and mtDNA replication during the cell cycle progression (Lin et al., 2003; Ling et al., 2004; Diaz and Moraes, 2008).

PGC-1α exhibits a tissue-enriched expression pattern and is highly inducible. This coactivator is enriched in tissues with a high-capacity mitochondrial system, such as brown fat, heart, and oxidative skeletal muscle fibers. PGC-1α is rapidly induced by cold exposure, short-term exercise and fasting, all conditions known to increase ATP and heat demand, and also by nitric oxide (Scarpulla, 2011).

PRC (PGC-1-related coactivator) has several domains homologous to PGC-1α and functional studies indicated that it is able to regulate mitochondrial function in a manner similar to PGC-1α. PRC interacts directly with NRF-1, promoting its activation, and can also activate the transcription of cyt c, a NRF-1 target, through the cooperation with other factors including CREB (Scarpulla, 2011).

The third member of the family, PGC-1β, shows a greater degree of homology to PGC-1α than PRC. The expression pattern of PGC-1β is very similar to that of PGC-1α, being enriched in heart and brown adipose tissue. This coactivator is induced by fasting, but not in response to cold exposure and is able to coordinate mitochondrial biogenesis inducing NRF-1 target genes (Lin et al., 2002).

NRF-1, NRF-2, ERRα are transcription factors that directly recognize specific sequences in the promoter of several nuclear encoded mitochondrial genes (Chau et al., 1992; Evans and Scarpulla, 1990; Schreiber et al., 2003). NRF-1 and NRF-2 may coordinate respiratory subunit expression by activating the mitochondrial transcription machinery, since they have been associated with the expression of many nuclear genes that encode subunits of the five OXPHOS complexes and mitochondrial transcription factor A (TFAM) and the two isoforms of mitochondrial transcription factor B (TFB2M) (Scarpulla, 2008). ERR-α, instead, has a direct role in activating NRF-1 and NRF-2 (Scarpulla, 2011).

In mitochondrial disorders, respiratory chain impairment is frequently associated with an increase in mitochondrial mass, a common cellular strategy to compensate for the energy defect. Some degree of mitochondrial proliferation in LHON patient's skeletal muscles is shown by the increased subsarcolemmal succinate dehydrogenase (SDH) activity (Carelli et al., 1998; Valentino et al., 2004; Carta et al., 2005). Moreover, an increased succinate-cytochrome c reductase activity, with normal complex III activity, has been reported in blood cells from LHON patients carrying the 11778/MT-ND4 mutation, suggesting again a nuclear compensatory effect (Yen et al., 1996). Increased mtDNA copy number has also been reported in blood cells from LHON affected individuals and asymptomatic mutation carriers harboring the 11778/MT-ND4 and 14484/MT-ND6 mutations (Yen et al., 2002; Nishioka et al., 2004). Finally, it has been recently demonstrated that an increase in mitochondrial biogenesis induced by β-estradiol is able to rescue the energetic defect in LHON cybrids (Giordano et al., 2011), supporting the relevance of an efficient mitochondrial biogenesis in LHON and providing a plausible explanation for male prevalence.

Recently, Moreno-Loshuertos et al. (2011) demonstrated that cells harboring a mutation in the mitochondrial tRNA(Ile) activate mitochondrial biogenesis as a compensatory response triggered by the increase in mitochondrial ROS production. The authors proposed that both the pathogenic mechanism consisting in the abnormal folding of the tRNA(Ile) combined with the existence of a compensatory mechanism can explain the penetrance pattern of this mutation (Moreno-Loshuertos et al., 2011). Thus, induction of mitochondrial biogenesis as a compensatory mechanism, which can be triggered by an increased ROS production, may also be important in LHON.

Therapy

LHON and DOA, as well as most of the neurodegenerative syndromes with optic atrophy are considered untreatable conditions; neither supportive therapies with supplements (vitamins, coenzyme Q, etc), nor corticosteroids have been proven to be effective (Sadun et al., 2011). Recent antioxidant molecules, such as idebenone and EPI-743, have been tested with some promising results (Klopstock et al., 2011; Carelli et al., 2011a; Sadun et al., 2012). Genetic approaches to correct the inherited defect have also been proposed and debated (Guy et al., 2002; Oca-Cossio et al., 2003; Perales-Clemente et al., 2011).

Pharmacological approach

The most promising recent results concern two molecules with antioxidant properties, which are believed to by-pass complex I by transferring electrons directly to complex III. The first is idebenone, a quinone analog of coenzyme Q (Geromel et al., 2002), which has been initially reported as possibly effective more than a decade ago, in a LHON case of Japanese ancestry who recovered visual acuity after treatment (Mashima et al., 1992). A few anecdotal reports followed with sometimes contrasting results and the mechanism of action of idebenone remained poorly understood (Carelli et al., 1998; Mashima et al., 2000; Barnils et al., 2007). In 2011, the results of the first controlled clinical trial with idebenone (Klopstock et al., 2011) were reported, side by side with a retrospective systematic evaluation of a large series of LHON treated patients compared with a similar cohort of matched untreated LHON cases (Carelli et al., 2011a). Both studies reached similar results documenting the partial effectiveness of idebenone in ameliorating the final outcome in LHON patients, mostly by increasing the rate of visual recovery after the acute phase (Newman, 2011; Sabet-Peyman et al., 2012). Recent studies on the mechanism of action revealed some details on how idebenone works in the cell, and in particular within mitochondria (Haefeli et al., 2011; Angebault et al., 2011; Giorgio et al., 2012). These studies suggest that to be effective and most importantly not toxic, idebenone should be maintained in the reduced state (Giorgio et al., 2012), a task that in some cell types appears to be carried out by NAD (P) H:quinone oxidoreductase 1 (NQO1) (Haefeli et al., 2011). In the reduced state idebenone provides electron transfer to complex III, which may allow reoxidation of NADH in complex I deficiencies.

A second molecule is α-Tocotrienol quinone, also coded as EPI-743, whose activity depends upon reversible 2e-redox-cycling, and exerts a potent anti-oxidant effect in vitro (Shrader et al., 2011). In a small open-label trial, EPI-743 arrested disease progression and reversed vision loss in all but 1 of 5 consecutively treated patients with LHON, thus showing promising results that need to be consolidated by a prospective randomized clinical trial (Sadun et al., 2012).

Much attention and investment is currently oriented on the activation of mitochondrial biogenesis as a partially effective compensatory strategy (Wenz et al., 2008; Viscomi et al., 2011). This may be sufficient for some mitochondrial disorders including the inherited optic neuropathies. Molecules that are being experimented in vitro and in animal models include some old drugs, such as bezafibrate and rosiglitazone, and others, which are not currently approved drugs, such as resveratrol and AICAR (5-aminoimidazole-4-carboxamide ribonucleoside). Similarly, we recently showed that estrogens ameliorate the biochemical cellular defects of LHON cybrids, again by activating mitochondrial biogenesis, thus suggesting the potential therapeutic use of estrogen-like molecules (Giordano et al., 2011). Clinical trials in human patients are warranted to understand if this will be a successful therapeutic strategy.

Gene therapy approach

The proposal of gene-therapy for mitochondrial diseases, including LHON for which the tissue affected is easily accessible, is mainly based on the approach of nuclear allotopic expression of a recoded and adequately engineered mtDNA gene (Gray et al., 1996; Manfredi et al., 2002; Guy et al., 2002). In the case of LHON, to express a wild-type version of the specific mtDNA-encoded ND subunit in the nucleus, first it needs to be recoded according to the genetic code of nuclear DNA. The recoded wild-type ND subunit is engineered to carry the mitochondrial import signal and is delivered by an adeno-associated virus (AAV) vector to the nucleus of RGCs. Thus, the now nuclear-encoded wild-type ND subunit, after being expressed in the cytoplasm, is targeted to the mitochondria where it is assumed to co-assemble in complex I, competing with the mitochondrial-encoded mutant ND subunit, possibly complementing the biochemical defect. Over the years, serious critiques have been presented on the efficiency and feasibility of this approach (Oca-Cossio et al., 2003; Perales-Clemente et al., 2011). However, successful results have been recently reported in an animal model of LHON treated by the allotopic gene-therapy approach, and there are plans to run clinical trials on LHON patients by this gene-therapy strategy (Ellouze et al., 2008; Lam et al., 2010).

Different approaches of gene therapy have been also presented. Experiments in an animal model of LHON were based on the xenotopic expression of an alternative oxidase (Saccharomyces cerevisiae single subunit NADH oxidase Ndi1), which can re-establish the electron flow without coupled proton translocation, thus missing the energy‐conserving function of complex I (Marella et al., 2010; Chadderton et al., 2012). Another approach was based on the activation of mitochondrial biogenesis by treating cultured cells with donor mtDNA complexed with recombinant human TFAM, which are taken up by cells and driven into mitochondria, ultimately improving their respiration (Iyer et al., 2012).

Finally, gene therapy for OPA1-related DOA patients is a feasible hypothesis, even if, at present there are a dearth of reports on this issue. However, contrary to mtDNA-based diseases, the techniques for gene replacement therapy are well developed for nuclear gene defects, and the recent successful trial with and AAV-mediated gene replacement for Leber congenital amaurosis provides a good example, which may be hopefully applied soon to DOA patients as well (Maguire et al., 2008).

Future directions and conclusions

Many complexities of mitochondrial optic neuropathies still remain poorly understood, mainly due to the lack of a true genetic animal model for mtDNA-based optic neuropathies, such as LHON. A number of animal models were created with the specific purpose of reproducing a mitochondrial optic neuropathy by biochemical manipulations (Zhang et al., 2002), or by ingenious engineering of gene expression (Qi et al., 2003; Ellouze et al., 2008; Marella et al., 2010). However, none of these models faithfully fulfills the real situation of LHON, a disease with homoplasmic mutant mtDNA that hits with incomplete penetrance only the RGC system, ultimately leading to optic atrophy. A few true mito-mouse models with mtDNA mutations exist, but unfortunately the eye and, in particular, the RGC system and optic nerve have not been properly investigated (Wallace, 2001; Nakada and Hayashi, 2011). Thus, the most important challenge for the future will be to firmly establish an animal model of LHON, with which to work out many of the unsolved questions and accelerate effective therapies.

Two different animal models for OPA1-related DOA currently exist, and are being investigated (Williams et al., 2011). Both animal models reproduced, with slightly different features, an age-dependent optic nerve dysfunction mimicking the slowly progressive pathology affecting human optic nerves in DOA patients with OPA1 mutations. Currently, the most provocative observation, which highlights the usefulness of animal models, concerns the precocious involvement of RGC dendrites, with dendritic pruning, retraction of mitochondria towards the soma, and mitochondrial fragmentation, preceding dendritic loss in animals already showing visual dysfunction (Williams et al., 2010, 2012). Thus, at a time-point for which there is still complete integrity of RGCs somata and axons, the first sign of pathology may be in the afferent synaptic connectivity of these cells, observed in the presence of abnormal mitochondrial distribution and fusion–fission balance.

In conclusion, the eye, and specifically the optic nerve, is an ideal system with which to observe and perhaps treat the consequences of mitochondrial form of neurodegeneration. The eye's particular vulnerability to mitochondrial diseases and its remarkable access should be exploited. Thus, the eye is a true “mito-window” on the brain.