Home > E. Pathology by systems > Nervous system > Peripheral nerves > amyotrophic lateral sclerosis

amyotrophic lateral sclerosis

Tuesday 23 September 2003

Definition: Amyotrophic lateral sclerosis (ALS) is a fatal adult-onset disease in which motor neurons in the brain and spinal cord degenerate by largely unknown mechanisms.

Amytrophic lateral sclerosis (ALS) is characterized by neuronal muscle atrophy (amyotrophy) and hyperreflexia due to loss of lower motor neurons in the anterior horns of the spinal cord and upper motor neurons that project in corticospinal tracts, respectively.

The disease affects men slightly more frequently than women and becomes clinically manifest in the fifth decade or later. Five per cent to 10% of cases are familial, mostly with autosomal-dominant inheritance.

Epidemiology

ALS is one of the most common neuromuscular diseases worldwide, and people of all races and ethnic backgrounds are affected. One to 2 people per 100,000 develop ALS each year.

ALS most commonly strikes people between 40 and 60 years of age, but younger and older people can also develop the disease. Men are affected slightly more often than women.

"Familial ALS" accounts for approximately 5%–10% of all ALS cases and is caused by genetic factors. Of these, approximately 1 in 10 are linked to a mutation in copper/zinc superoxide dismutase (SOD1), an enzyme responsible for scavenging free radicals.

Although the incidence of ALS is thought to be regionally uniform, there are three regions in the West Pacific where there has in the past been an elevated occurrence of ALS. This seems to be declining in recent decades.

The largest is the area of Guam inhabited by the Chamorro people, who have historically had a high incidence (as much as 143 cases per 100,000 people per year) of a condition called Lytico-Bodig disease which is a combination of ALS, Parkinsonism, and dementia.

Two more areas of increased incidence are the Kii peninsula of Japan and West Papua.

Clinical synopsis

Early symptoms include asymmetric weakness of the hands, manifested as dropping objects and difficulty in performing fine motor tasks, and cramping and spasticity of the arms and legs.

As the disease progresses, muscle strength and bulk diminish, and involuntary contractions of individual motor units, termed fasciculations, occur.

The disease eventually involves the respiratory muscles, leading to recurrent bouts of pulmonary infection. The severity of involvement of the upper and lower motor neurons is variable; the term progressive muscular atrophy applies to those relatively uncommon cases in which lower motor neuron involvement predominates.

In some patients, degeneration of the lower brainstem cranial motor nuclei occurs early and progresses rapidly, a pattern referred to as progressive bulbar palsy or bulbar amyotrophic lateral sclerosis. In these individuals, abnormalities of deglutition and phonation dominate, and the clinical course is inexorable during a 1- or 2-year period.

Familial cases develop symptoms earlier than most sporadic cases do, but the clinical course is comparable.

Familial ALS

ALS is familial (FALS) in 10% of cases, and the inheritance is usually dominant, with variable penetrance. Mutations in copper/zinc super oxide dismutase (SOD1) are found in 20% of familial and 3% of sporadic ALS cases.

Although most cases of amyotrophic lateral sclerosis are sporadic, familial cases have been identified.

In approximately 20 percent of familial cases of amyotrophic lateral sclerosis, there are mutations in the gene for cytoplasmic superoxide dismutase type 1 (SOD1).

Moreover, deposits of SOD1 in the central nervous system have been found in both sporadic and familial cases of amyotrophic lateral sclerosis.

Although in some cases abnormal collections of neurofilaments have been seen in degenerating motor neurons, no familial cases have been shown to be due to mutations in neurofilament genes.

Pathogenesis

The etiology and pathogenesis of amyotrophic lateral sclerosis are unknown. For a subset of the familial cases, the genetic locus has been mapped to the copper-zinc superoxide dismutase gene (SOD1) on chromosome 21.

A wide variety of missense mutations have been identified that appear to generate an adverse gain-of-function phenotype. Among the mutations, the A4V mutation is the most common (approaching 50% of cases), is associated with a rapid course, and rarely has upper motor neuron signs.

A recessive locus on chromosome 2 has been mapped to a gene encoding a protein termed alsin that has structural homology to proteins involved in GTPase regulation.

Other genetic loci for ALS have been mapped but not yet cloned. There is also evidence of roles of glutamate toxicity and protein nitration in the development of ALS pathology. The basis for the selective involvement of motor neurons remains uncertain.

Morphology

On macroscopic examination, the anterior roots of the spinal cord are thin; the precentral gyrus may be atrophic in especially severe cases.

Microscopic examination demonstrates a reduction in the number of anterior horn neurons throughout the length of the spinal cord with associated reactive gliosis and loss of anterior root myelinated fibers. Similar findings are found with involvement of the hypoglossal, ambiguus, and motor trigeminal cranial nerve nuclei.

Remaining neurons often contain Bunina bodies: PAS-positive cytoplasmic inclusions that appear to be remnant of autophagic vacuoles. Skeletal muscles innervated by the degenerated lower motor neurons show neurogenic atrophy.

Destruction of the upper motor neurons leads to degeneration of myelin in the corticospinal tracts, resulting in pale staining that is particularly evident at the lower segmental levels but traceable throughout the corticospinal system with special studies.

Types

familial amyotrophic lateral sclerosis (10%)
sporadic amyotrophic lateral sclerosis

Typical ALS

ALS1 (adult) AD 21q22.1 Superoxide dismutase 1, SOD1 (1993)
ALS3 (adult) AD 18q21 ?
ALS6 (adult) AD 16q12 ?
ALS7 (adult) AD 20p13 (1 family) ?

ALS with Dementia

ALS with frontotemproal dementia (ALSFTDs)

FTD-ALS
ALSFTD1 at 9q21-22|
ALSFTD2
TARDP-associated ALS-FTDU (adult) at 1q32
TAU-associated ALS-FTDU at 17q21.1

Atypical ALS

ALS8 (adult) AD 20q13.3 VAMP associated protein B (VAPB) (2004)
Progressive LMN (adult) AD 2p13 Dynactin p150 subunit (2004)

Other Motor Neuron Diseases, referred to as ALS

ALS2 (juvenile) AR 2q33 ALS2/Alsin (2001)
ALS4 (juvenile) AD 9q34 Senataxin (2004)
ALS5 (juvenile) AR 15q15.1-21.1 ?

By loci

ALS1	21q22.1	MIM.105400	SOD1	superoxyde dismutase-1	MIM.147450
ALS2	2q33	MIM.205100	ALS2	alsin	MIM.606352
ALS3	18q21	MIM.606640	-	-	-
ALS4	9q34	MIM.602433	SETX	senataxin	MIM.608465
ALS5	15q15.1-q21.1	MIM.602099	-	-	-
ALS6	16q12	MIM.608030	-	-	-
ALS7	20p13	MIM.608031	-	-	-
ALS8	20q13.3	MIM.608627	VAPB	vesicle-associated membrane protein	MIM. 605704
ALS9	14q11	MIM.611895	ANG	angiogenin	MIM.105850

ALS genes

SOD1

ALS2

SETX

VAPB

ANG

DCTN1

MAPT

TARDBP

Etiology of familial ALS

ALS1: 15 to 20% of cases of familial amyotrophic lateral sclerosis (FALS) and 2% of total cases) are associated with mutations in the cytosolic Cu/Zn superoxide dismutase-1 gene (SOD1) (MIM.147450) on chromosome 21q22.1.

Although most cases of SOD1-related familial ALS follow autosomal dominant inheritance, autosomal recessive inheritance has been rarely reported.

Susceptibility to ALS has been associated with mutations in other genes, including deletions or insertions in the gene encoding the heavy neurofilament subunit (NEFH) (MIM.162230); deletions in the gene encoding peripherin (PRPH) (MIM.170710); and mutations in the dynactin gene (DCTN1) (MIM.601143).

ALS6: 16q12.1-16q12.2 locus (12858291, 12840784, 12830400)

ALS and neuronal intermediate filaments

The genes encoding the three neurofilament subunits have long been suspected as causative for ALS because of their link with motor neuron pathology in mice and humans.

Initial evidence fueling this suspicion was the realization that aberrant neurofilament accumulations are a pathological hallmark of both familial and sporadic ALS.

In addition, mice with increased levels of wild-type NF-H or NF-L subunits develop age-dependent motor neuron pathology, while expression of a point mutation in NF-L at levels corresponding to that expected for dominantly inherited disease produces fatal, progressive paralysis.

Despite this, exhaustive screening for mutations in the three neurofilament genes in patient DNAs failed to yield conclusive linkage to either sporadic or familial ALS patients, although dominant point mutations in NF-L have been linked to a milder motor neuron disease, Charcot-Marie-Tooth (CMT) disease.

Nevertheless, in-frame insertions or deletions within the normal array of 44 or 45 KSP repeats in the tail domain of the NF-H subunit have been reported in 1% of sporadic ALS patients. Collectively, it is likely that these neurofilament variants are risk factors for ALS and/or a primary event with low penetrance.

A final potential contribution of neurofilaments to ALS comes from errors in the expression of an additional intermediate filament subunit, peripherin (PRPH1). An assembly-disrupting frameshift mutation has been reported in one sporadic ALS case. A proposal for the involvement of a peripherin variant, in which an intronic sequence is retained in the final mRNA that is produced during the disease course, may also contribute to human disease, albeit deletion of the peripherin gene has no affect on disease course in mice having an ALS-causing SOD1 mutation.

Motor Neuron Death from Toxicity of Mutant SOD1 Not Loss of Dismutase Activity

SOD1 is an abundant, ubiquitously expressed cytosolic enzyme. Since the known activity of SOD1 is to dismutate (or convert) superoxide, a natural byproduct of respiration, to water or hydrogen peroxide, the first proposed pathogenic mechanism was the loss or diminution of this detoxifying activity.

However, animals expressing dismutase active (hSOD1G37R [Wong et al., 1995] and hSOD1G93A [Gurney et al., 1994 and Howland et al., 2002]) as well as inactive (hSOD1G85R [Bruijn et al., 1997], mSOD1G86R [Ripps et al., 1995], hSOD1G127X [Jonsson et al., 2004], hSOD1Quad [Wang et al., 2003], hSOD1H46R [Nagai et al., 2001]) forms of the enzyme develop comparable disease pathologies similar to that seen in patients, including motor neuron synapse retraction (Fischer et al., 2004, Frey et al., 2000 and Pun et al., 2006), mitochondrial alterations (Higgins et al., 2003, Kong and Xu, 1998, Liu et al., 2004 and Wong et al., 1995), and microglial activation (Boillée et al., 2006, Engelhardt and Appel, 1990, Hall et al., 1998 and Henkel et al., 2006).

Furthermore, SOD1 gene deletion in mice does not lead to motor neuron disease (Reaume et al., 1996). In addition, deletion of the endogenous mouse SOD1 in mice expressing dismutase inactive hSOD1G85R does not affect disease course (Bruijn et al., 1998). Mice expressing high levels of wild-type human SOD1 (hSOD1WT) transgene are healthy (Gurney et al., 1994 and Wong et al., 1995). Indeed, increased hSOD1WT, accompanied by chronic elevation of dismutase activity, has either no effect on disease (Bruijn et al., 1998) or accelerates it (Deng et al., 2006 and Jaarsma et al., 2000). Taken together, these experiments have argued heavily against the simple loss-of-function hypothesis.

SOD1 activity is dependent on a catalytic copper. Since free copper is highly reactive and toxic, it must be loaded onto SOD1 by a copper chaperone for SOD1 (CCS) and subsequently held in place by a conserved disulfide bond whose formation is catalyzed by CCS.

Since CCS is abundantly expressed in motor neurons and motor neurons of CCS-deleted mice have an increased sensitivity to axotomy-induced death, it was postulated that inefficient incorporation of copper into SOD1 and/or a decreased shielding of copper (due to changes in SOD1 structure as a result of mutation) could provide an opportunity for aberrant copper-dependant oxidative chemistry, yielding cellular damage and motor neuron degeneration.

This hypothesis was tested using mice in which the incorporation of copper into mutant SOD1 was significantly reduced by disruption of the CCS gene. This did not slow disease course from either dismutase active or inactive mutants (Subramaniam et al., 2002). Furthermore, mice in which all four copper-binding histidines have been eliminated (hSOD1Quad; resulting in dismutase inactive SOD1) still develop typical ALS-like motor neuron disease (Wang et al., 2003).

Taken together, it is clear that the toxic property (or properties) acquired as a result of SOD1 mutation (1) are independent of dismutase and CCS activities and (2) can be generated without catalysis by the active site copper.

Misfolded SOD1 as a Common Feature of ALS-Causing Mutations

Protein aggregates are a pathological hallmark of many different neurodegenerative disorders including Alzheimer disease, Parkinson disease, Huntington diseases and ALS.

Cytoplasmic protein aggregates are observed in both sporadic and familial ALS cases as well as in mutant SOD1 transgenic mice. That said, use of the term “aggregate” is imprecise. In some instances, it refers to abnormal accumulations of intermediate filaments, including neurofilaments and peripherin, as detected by immunostaining of spinal cord tissue.

It has also been used to define accumulations of detergent-insoluble forms of proteins, including SOD1, that are detected by immunoblotting of filter-trappable material, as well as small SOD1- or ubiquitin-positive fibrillar inclusions in spinal cord sections.

Detergent-insoluble species are detectable only in affected tissues of mutant SOD1 mice and are most prominent at symptomatic stages. A propensity to form aggregates following synthesis of mutant SOD1 in primary cells is selective to motor neurons, as aggregates were not produced in dorsal root ganglion (DRG) or hippocampal neurons expressing similar levels of mutants. The most misfolded unstable SOD1 mutants (with the shortest in vivo half-lives) are most prone to aggregation.

A very curious and unexplained aspect of disease is that the very unstable mutant hSOD1-A4V, which provokes a very aggressive disease course in humans, neither induces aggregates nor ALS-like disease in mice, except in the context of high levels of hSOD1-WT.

Similarly, increased hSOD1-WT accelerated disease from a second unstable mutant hSOD1-L126Z. In both cases, accelerated toxicity may be mediated by stabilization of the mutant via heterodimerization with hSOD1WT protein. Indeed, in both cases, increased levels of hSOD1WT generated detergent-insoluble forms of SOD1 that were not seen in spinal cord extracts of hSOD1A4V or hSOD1L126Z mice alone.

Are SOD1 Aggregates Toxic?

Regardless of their definition or composition, the contribution of aggregates to motor neuron toxicity remains to be established. Indeed, it is unknown whether any of these aggregates are indeed toxic.

Proposals include aggregate-mediated inhibition of the proteasome machinery, decreased chaperone activity, deregulation of organelle function including Golgi, endoplasmic reticulum, and mitochondria, and axonal transport defects possibly linked to aberrant accumulations of intermediate filaments.

Aggregates found in ALS patients, as well as mouse models, contain ubiquitin, a protein adduct which typically targets proteins for disposal via the proteasome.

Misaccumulation of ubiquitinated, misfolded proteins might adversely affect the proteasome machinery and impair normal protein degradation.

Indeed, the degradation of SOD1 itself and aggregates of SOD1 is proteasome mediated. While proteasome malfunction has been implicated in motor neuron death, it is not yet established if it is a cause or consequence of aggregate formation. Contradictory results have been reported.

In hSOD1-G93A mice which accumulate mutant protein to high levels, proteasome activity is downregulated in the lumbar spinal cord well before the development of symptoms, while in a different line of mice with lower accumulated levels of the same mutant, increased proteasome activity in the spinal cord has been reported at symptomatic stages.

Protein misfolding

Protein misfolding is a probable initiator of aggregate formation. Thus, many efforts have examined the protein folding chaperone machinery, the heat-shock proteins (HSPs), in the context of disease.

In spinal cord extracts of presymptomatic ALS mice, an overall decrease of chaperone activity has been reported which persists throughout disease course, and multiple recombinant SOD1 mutants inhibit chaperone function in vitro.

Paradoxically, some HSPs, such as B-crystallin and Hsp27 are elevated in the spinal cords of hSOD1-G37R and hSOD1-G93A mice, but Hsp27 is predominantly present in glial cells in late disease stages.

Hsp70, Hsp40, and Hsp90 are also elevated but only in the spinal cords of hSOD1G85R mice.

Motor neurons have a high threshold for induction of the stress response, at least in vitro, which may contribute to their selective vulnerability in ALS.

Mutant SOD1-mediated depletion of HSPs is a plausible possibility given that Hsp70 and Hsp25 preferentially bind mutant SOD1.

Expression of several different HSPs (Hsp70, Hsp40, Hsp27) in cultured cells and primary motor neurons decreases aggregate content and apoptosis and improves axonal outgrowth.

Unfortunately, applying this strategy in vivo by increased expression of Hsp70 in four different mutant SOD1 mouse lines did not ameliorate disease or pathology.

Induction of a broader stress response extended life span in a small cohort of hSOD1G93A mice after treatment with arimoclomol, a drug which induces the phosphorylation-mediated activation of the HSP inducing factor HSF-1, thereby leading to increased levels of Hsp70 and Hsp90 in spinal cords.

Interestingly, missense mutations in the gene encoding Hsp27 have been identified in a number of families with distal hereditary motor neuropathies and CMT neuropathies.

Mitochondrial Dysfunction as a Proximal Cause of Motor Neuron Death in ALS

A common feature of many neurodegenerative diseases is damage to mitochondria that contributes to the degenerative phenotype.

Mitochondria were implicated as a possible target for toxicity in ALS by histopathological observations of vacuolated and dilated mitochondria with disorganized cristae and membranes in the motor neurons (and muscle) of both sporadic and familial ALS patients.

Mitochondrial defects have been reported in spinal cords and muscle biopsies of patients ranging from histopathological observations to impaired mitochondrial respiration and increased levels of uncoupling proteins. However, since all patients tested were obligatorily symptomatic, it is difficult to determine if these defects are primary or secondary to muscle atrophy.

Mitochondrial abnormalities, especially multimembrane-containing vacuoles, are observed in the motor neurons of mice that develop disease from accumulation of dismutase active mutants (hSOD1-G37R and hSOD1-G93A mice) at very early, presymptomatic stages (prior to any detectable motor neuron loss or other observable damage.

It has been proposed that vacuole formation is due to expansion of the mitochondrial intermembrane space and consequent distention of mitochondrial membranes.

However, vacuoles inferred to derive from degeneration of mitochondria have also been described in mice with very high accumulated levels of hSOD1WT protein which do not develop disease and are absent in mice that develop disease from dismutase inactive mutants. Thus, vacuolation cannot be a primary/common pathogenic mechanism for motor neuron degeneration.

A proportion of the predominantly cytosolic SOD1 localizes to mitochondria in certain contexts. In both rodent models and patient samples, mutant SOD1 is present in fractions enriched for mitochondria derived from affected, but not unaffected tissues.

Mitochondrial localization of SOD1 has been confirmed by electron microscopy in both isolated mitochondria and motor neurons in situ.

SOD1 mutants that cause disease at the lowest accumulated levels (hSOD1G85R and hSOD1G127X; all dismutase inactive) have the highest relative proportions that are mitochondrially associated (Liu et al., 2004). There is disagreement in defining the submitochondrial compartment(s) with which SOD1 is localized (or potentially aggregated).

Mutant SOD1 has been reported in both the intermembrane space and the matrix, as well as both the inner and outer membranes of spinal cord and brain mitochondria.

While these apparent contradictions remain to be resolved, all reports agree that SOD1 mutant proteins of divergent biochemical characteristics localize to mitochondria, consistent with a common contribution to pathogenesis.

Key questions are unresolved. Is there some intrinsic feature of spinal cord mitochondria which permits SOD1 association and/or increases mitochondrial vulnerability to mutant SOD1? Alternatively, is it the cytoplasmic components of chaperone or proteasome activities which confers some selectivity and facilitates mitochondrial association ?

In which cell type(s), if any, does mitochondrial association of the mutant play a central role in pathogenesis? Of relevance here, the endogenous wild-type SOD1 protein is largely excluded from spinal cord mitochondria, but all human SOD1 mutants examined to date are mitochondrially associated.

The same human mutant proteins are largely excluded from similar preparations of liver mitochondria despite a proportion of endogenous mouse SOD1 localized to the intermembrane space of those mitochondria. This discordance highlights that not all mitochondria are created equally, as supported by recent proteomic efforts profiling mitochondrial heterogeneity across tissues.

How mutant SOD1 affects mitochondrial function is not yet clear, but differences in protein composition between mutant and nonmutant mitochondrial populations have been reported.

It is important to note that the mitochondrial genome itself encodes for only thirteen of the estimated 1500 proteins required for function. The majority of mitochondrial proteins are nuclear encoded and thus are synthesized in the cytosol and subsequently transported into mitochondria.

Import of the majority of mitochondrial proteins is accomplished by membrane-spanning multisubunit translocators of the outer and inner membrane (TOM and TIM, respectively). Mutant SOD1 associated with or aggregated onto the mitochondrial surface could impede the import machinery.

Many aspects of the electron transport chain, which generates ATP via oxidative phosphorylation, have been reported as either unchanged or altered, although there is little or no consensus as to whether these activities are accentuated or attenuated.

For example, ATP synthesis has been reported as unchanged in aged hSOD1-G85R mice (Damiano et al., 2006) or depleted in late symptomatic hSOD1-G93A mice (Mattiazzi et al., 2002), or ATP levels are depleted in presymptomatic mice (Browne et al., 2006).

Creatine, which extended survival in hSOD1-G93A mice by alleviating presumed energy deficits (Browne et al., 2006 and Klivenyi et al., 1999), was ineffective in human clinical trials (Groeneveld et al., 2003 and Shefner et al., 2004).

Complex I activity has been reported as elevated (Bowling et al., 1993 and Browne et al., 1998), reduced (Jung et al., 2002, Kirkinezos et al., 2005 and Mattiazzi et al., 2002), and unchanged (Damiano et al., 2006).

Complex IV activity is reportedly reduced, but only at very late disease stages (Kirkinezos et al., 2005 I.G. Kirkinezos, S.R. Bacman, D. Hernandez, J. Oca-Cossio, L.J. Arias, M.A. Perez-Pinzon, W.G. Bradley and C.T. Moraes, Cytochrome c association with the inner mitochondrial membrane is impaired in the CNS of G93A-SOD1 mice, J. Neurosci. 25 (2005), pp. 164–172. Full Text via CrossRef | View Record in Scopus | Cited By in Scopus (26)Kirkinezos et al., 2005 and Mattiazzi et al., 2002).

Early impairment in mitochondrial calcium-buffering capacity in mutant SOD1 spinal cord prior to symptoms and only in disease-relevant tissues in two different mutant SOD1 models (Damiano et al., 2006) is perhaps the most persuasive of the efforts, since mitochondrial control of calcium homeostasis is fully compatible with excitotoxic mechanisms that may be central to neuronal damage.

Mutant SOD1 is preferentially imported into or deposited onto mitochondria in affected tissues. The mechanisms by which mutant protein might damage mitochondrial function are drawn.

Mutant SOD1 may interfere with the elements of the electron transport chain, thereby disrupting ATP-generating oxidative phosphorylation. Mutant SOD1 may also disrupt mechanisms by which mitochondria buffer cytosolic calcium levels. Mutant SOD1 aggregates may interfere with components of mitochondrial-dependent apoptotic machinery, such as Bcl-2, thereby triggering premature activation of an apoptotic cascade including cytochrome c release into the cytosol.

Mutant SOD1 may indirectly affect similar pathways linked to mitochondria by physically blocking the protein import machines, TOM and TIM.

Oxidative damage incurred by various mitochondrial proteins may also contribute to overall mitochondrial dysfunction. Collectively, these mechanisms (or a combination thereof) are predicted to disturb cellular homeostasis (within glial and/or motor neurons), ultimately triggering motor neuron death.

Four stages are defined (normal, early phase, symptomatic, and end stage). Toxicity is non-cell-autonomous, produced by a combination of damage incurred directly within motor neurons that is central to disease initiation and damage within nonneuronal neighbors, including astroctyes and microglia, whose actions amplify the initial damage and drive disease progression and spread.

Selective vulnerability of motor neurons to ubiquitously expressed mutant SOD1 is determined by the unique functional properties of motor neurons (e.g., they are very large cells with large biosynthetic loads, high rates of firing, and respond to glutamate inputs) and damage to their supporting cells in the neighborhood.

Mitochondria: The Gatekeepers of Cell Death

Mitochondria are the gatekeepers of apoptosis, with opening of the permeability transition pore and release of cytochrome c central to the cascade of caspase activation.

A final apoptotic mechanism is probably a central aspect of neuronal death, as activation of the executioner caspase-3 is found in mouse models contemporaneous with neuronal death (Li et al., 2000 and Pasinelli et al., 2000) and lowered levels of the antiapoptotic Bcl-2 have been reported in spinal cord motor neurons of ALS patients (Ekegren et al., 1999 and Mu et al., 1996) and mutant SOD1 mice (Gonzalez de Aguilar et al., 2000 and Vukosavic et al., 2000).

Intriguingly, it has recently been demonstrated that caspase-3 activation in glial cells proteolytically inactivates the glutamate transporter EAAT2 (Boston-Howes et al., 2006).

Since EAAT2 is selectively lost during the disease process and is considered to be a main participant in excitotoxicity (which can trigger additional mitochondrial dysfunction), this provides an attractive mechanism by which to unite mitochondrial dysfunction with excitotoxicity (see later).

An alternate mechanism, in which SOD1 is proposed to interact with Bcl-2 and possibly interferes with Bcl-2 antiapoptotic activity (Pasinelli et al., 2004), is attractive but has recently been disputed (Gould et al., 2006).

However, both constitutively increased expression of Bcl-2 (Kostic et al., 1997) and virally driven overexpression of Bcl-2 in motor neurons (Azzouz et al., 2000) protect motor neurons, delaying disease onset but not progression.

Bcl-2 is unable to similarly rescue motor neuron death in other models, such as wobbler and transgenic mice expressing a point mutation in the NF-L gene (Coulpier et al., 1996 and Houseweart and Cleveland, 1999).

Moreover, consistent with a similar outcome in which intracerebroventricular (ICV) infusion of a broad spectrum caspase inhibitor (Li et al., 2000 M. Li, V.O. Ona, C. Guegan, M. Chen, V. Jackson-Lewis, L.J. Andrews, A.J. Olszewski, P.E. Stieg, J.P. Lee, S. Przedborski and R.M. Friedlander, Functional role of caspase-1 and caspase-3 in an ALS transgenic mouse model, Science 288 (2000), pp. 335–339. Full Text via CrossRef | View Record in Scopus | Cited By in Scopus (344)Li et al., 2000) slowed mutant SOD1-mediated neuronal death, disease onset was also delayed in the absence of Bax, but progression was unaffected (Gould et al., 2006), casting doubt on the utility of antiapoptotic therapies.

Loss of the Neuromuscular Synapse as the First Cellular Phenotype in ALS

The axon is the only means by which a motor neuron communicates with its downstream target muscle, both sending output information and receiving signaling inputs from the muscle and other cells. Connection to the muscle at the neuromuscular junction is lost in ALS mouse models long before motor neuron degeneration or death and the initiation of symptoms (Fischer et al., 2004, Frey et al., 2000, Kong and Xu, 1998 and Pun et al., 2006).

It has been long established in humans (Kawamura et al., 1981) and in mice (Bruijn et al., 1997 and Wong et al., 1995) that it is the motor neurons that generate large caliber (>5 μm) myelinated axons that are selectively vulnerable in ALS.

This important distinction has now been refined with the demonstration that different types of motor neurons, innervating different subsets of muscle fibers, have different susceptibilities to mutant SOD1 toxicity.

Specifically, in two different mutant SOD1 models, the fast-fatiguable motor neurons were shown to be affected first, with denervation of their target muscles well before symptoms.

Retraction of the fast-fatigue-resistant motor neurons followed next, with the slow type partially resistant to mutant SOD1 and actually attempting to reinnervate previously denervated regions (Frey et al., 2000 and Pun et al., 2006).

A spontaneous dominant mouse mutant Wlds (Wallerian degeneration slow) significantly delays the degeneration of axonal segments located distal of an axonal cut (axotomy; Glass et al., 1993 and Perry et al., 1990).

This is due to a unique fusion protein containing the first 70 residues of the ubiquitination factor UFD2/E4 joined to mononucleotide adenylyltransferase (Nmnat), an enzyme that facilitates nicotinamide adenine dinucleotide (NAD) synthesis (Conforti et al., 2000 and Mack et al., 2001).

Despite conferring robust axonal protection in neuropathy mouse models, including the peripheral neuropathy model (pmn/pmn; Ferri et al., 2003) and a model of myelin-related axonopathy (P0-deficient; Samsam et al., 2003), Wlds failed to provide a benefit in three different mutant SOD1 models (Fischer et al., 2005 and Vande Velde et al., 2004).

Damage within the Axon - Compromised Axonal Transport in ALS

Motor neurons are among the most asymmetric cells in nature, extending axons in humans that can be more than a meter in length. Alterations in axonal structure are well documented in both patients (Hirano et al., 1984a, Hirano et al., 1984b and Kawamura et al., 1981) and mutant SOD1 mice (Bruijn et al., 1998, Kong and Xu, 1998 and Wong et al., 1995), especially the inappropriate misaccumulation of neurofilaments, the most abundant structural components of large myelinated axons.

Neurofilaments are obligate heteropolymers of NF light (NF-L), NF medium (NF-M), and NF heavy (NF-H) subunits and interestingly, transgenic mice expressing a point mutation in NF-L develop motor neuron disease (Lee et al., 1994).

Neurofilament accumulations are seen early in mutant SOD1 mice (Kong and Xu, 1998). Removal of all axonal neurofilaments by deletion of NF-L substantially prolonged survival of hSOD1G85R and hSOD1G37R mice (Nguyen et al., 2001 and Williamson et al., 1998), as did removal of most axonal neurofilaments by excessive levels of NF-H (Couillard-Despres et al., 1998, Kong and Xu, 2000 and Nguyen et al., 2001).

Since neurofilament-dependent slowing of slow axonal transport is itself directly dependent on the phosphorylation state of neurofilament tail domains (Ackerley et al., 2003), the increased survival benefit likely derives primarily from reducing the neurofilament-dependent burden on the axonal transport machinery.

This was demonstrated to be the case by use of gene replacement to remove the phosphorylated “tail” domains of NF-M and NF-H that normally provide intra-axonal crosslinking of adjacent filaments.

Absence of those tail domains sharply slows SOD1 mutant-induced disease (Lobsiger et al., 2005). Consistent with this, a presymptomatic deficit of slow axonal transport has been described in mutant SOD1 mice (Ackerley et al., 2003, Williamson and Cleveland, 1999 and Zhang et al., 1997).

Additional evidence for a link between transport and motor neuron disease arose from the discovery of a mutation in the p150Glued subunit of dynactin in a family affected with a slowly progressive, autosomal dominant form of lower motor neuron disease (Puls et al., 2003). p150Glued is responsible for providing processivity by bridging between microtubules and cytoplasmic dynein (Vaughan and Vallee, 1995 and Waterman-Storer et al., 1995), a motor powering retrograde axonal transport along microtubules (Goldstein and Yang, 2000).

Mice heterozygous for either of two point mutations in dynein have mispositioned hindlimbs, a phenotype reflected in the names of the mutations, Legs at odd angles (Loa) and Cramping1 (Cra1; Hafezparast et al., 2003). Surprisingly, introducing either the Loa (Kieran et al., 2005) or Cra1 (Teuchert et al., 2006) mutation to hSOD1G93A mice significantly improves survival. This unexpected amelioration of SOD1 mutant toxicity by a mutation expected to alter retrograde axonal transport is indeed a conundrum that remains to be resolved.

Further genetic evidence implicating axonal transport defects as a bona fide pathogenic mechanism in motor neuron degeneration stem from additional mutations in multiple genes relevant to transport in both human disease and mouse models.

These include loss of the kinesin motor protein KIF1Bβ as causative for CMT type 2A (Zhao et al., 2001), loss of KIF5A in hereditary spastic paraplegia (Reid et al., 2002), and loss of KIF21A in a rare disorder affecting the oculomotor nerve (Yamada et al., 2003).

Mutations in genes involved in transport of membranous vesicles include the ALS8 gene VAPB (Nishimura et al., 2004b), Rab7 in CMT type 2B (Verhoeven et al., 2003), and Vps54 (vacuolar-vesicular protein sorting 54) which is responsible for cervical motor neuron degeneration in the wobbler mouse (Schmitt-John et al., 2005).

The Neighborhood Matters: Non-Cell-Autonomous Death of Motor Neurons

Although progressive paralysis in ALS arises from degeneration and death of motor neurons, evidence from several directions has converged to demonstrate that toxicity is non-cell-autonomous. That is, toxicity to motor neurons derives from damage developed within cell types beyond the motor neurons.

Consistent with this, the known inherited forms are caused by mutations in genes that are ubiquitously expressed (SOD1 and VAPB) or expressed in multiple cells types (VEGF and ANG).

The initial evidence that damage within more than one cell type was required for disease came from expression of mutant SOD1 only within motor neurons (Lino et al., 2002 and Pramatarova et al., 2001) or astrocytes (Gong et al., 2000).

None of these efforts produced motor neuron degeneration or death (albeit the expression of mutant SOD1 selectively within motor neurons might have been at levels too low to initiate disease).

More definitive evidence emerged from construction and analysis of chimeric mice that were mixtures of hSOD1 mutant-expressing cells and normal cells (Clement et al., 2003).

Some animals with only a small minority of normal cells escaped disease, despite having high proportions of mutant motor neurons. These efforts conclusively demonstrated that expression of mutant SOD1 within individual motor neurons, even at levels that cause early onset, rapidly progressing disease when expressed ubiquitously, is not sufficient to provoke cell-autonomous degeneration or death of individual motor neurons of comparable ages.

Construction and analysis of mice carrying a deletable (“floxed”) mutant SOD1 gene that can be excised by the action of the Cre recombinase has identified differential contribution of mutant damage within motor neurons and nonneuronal neighbors in triggering disease onset and progression (Boillée et al., 2006).

Excision of the floxed mutant SOD1 gene exclusively within motor neurons (by action of Cre recombinase expressed under control of the motor-neuron-specific promoter Islet-1) extended survival by slowing onset and an early phase of disease progression.

Similar findings by altering components expressed only within neurons (i.e., gene disruption to remove all neurofilaments [Williamson et al., 1998] or gene replacement to alter axonal structure and volume after elimination of the tail domains of the NF-M and NF-H subunits [Lobsiger et al., 2005]) provide extended survival of SOD1 mutant mice but only by slowing disease onset.

In contrast, diminishing mutant SOD1 levels within microglia and peripheral macrophages (using a Cre transgene with a CD11b promoter that is expressed only within the microglia and peripheral macrophage lineages) had little effect on early disease but sharply slowed later disease progression, extending overall survival by 99 days.

Similarly, in a complementary approach the entire myeloid lineage was replaced by transplantation of normal bone marrow cells into SOD1 mutant mice that were themselves unable to synthesize their own myeloid cells due to deletion of the transcription factor PU.1 (Beers et al., 2006).

Transplantation at birth with hSOD1G93A mutant-expressing myeloid cells (which populate both the CNS and the periphery) produced onset and survival typical of the hSOD1G93A mutant line. Replacement of the microglial, monocyte, and macrophage lineages with normal cells had no effect on disease onset but slowed disease progression after onset.

Thus, both approaches demonstrated that mutant SOD1 within macrophages/microglial cells accelerates disease progression, while mutant action within the motor neurons is a primary determinant of onset and early disease. Importantly, introducing mutant SOD1-expressing microglial cells into control animals did not give rise to motor neuron disease, demonstrating that mutant SOD1-expressing macrophages/microglial cells themselves are not sufficient to cause motor neuron death.

It should be emphasized that a contribution of peripheral macrophages to accelerating disease progression cannot be excluded. Indeed, a proportion of macrophages/microglial cells in the spinal cord of mutant SOD1 mice have been shown to enter from the periphery during the course of disease in mice (albeit this contribution seems minor [Solomon et al., 2006]).

There is disagreement whether replacing all of the peripheral, but only a few of the spinal cord, myeloid derived cells by bone marrow transplantation following irradiation is beneficial in slowing disease (Corti et al., 2004 and Solomon et al., 2006). However, there is no uncertainty that replacing both the peripheral and CNS microglial pools slows disease progression after onset (Beers et al., 2006).

Contributions to toxicity developed within astrocytes, the myelinating Schwann cells, or the target muscles have not yet been established, although the “floxed” mutant transgene approach is well suited for such tests.

Targets for Therapy: Microglial Cells as Determinants of Disease Progression

Microglial cells, the macrophages of the CNS, have been long suspected as central components in neurodegenerative diseases where their role may include secretion of trophic or toxic molecules.

Their role in neuronal degeneration during development is well established (Marin-Teva et al., 2004). In ALS, microglial activation has been described in the brain and spinal cord of patients (Engelhardt and Appel, 1990, Henkel et al., 2004, Kawamata et al., 1992, McGeer et al., 1991, Troost et al., 1993 and Turner et al., 2004) and in the spinal cord of different mutant SOD1 mouse models (Hall et al., 1998, Henkel et al., 2006 and Kriz et al., 2002).

The microglial reactivity is initiated before motor neuron loss (Henkel et al., 2006). In addition, increased numbers of dendritic cells, potent antigen-presenting cells implicated in the immune response, have been reported in spinal cords of ALS patients and hSOD1G93A mice during the disease course (Henkel et al., 2006 and Henkel et al., 2004).

Several groups have tried to treat motor neuron disease by using minocycline, a tetracycline derivative previously shown to inhibit microglial activation (Yrjanheikki et al., 1999).

Minocycline was potent in increasing survival of ALS mice and reduced microglial activation (Kriz et al., 2002, Van Den Bosch et al., 2002 and Zhu et al., 2002). The primary effect of minocycline treatment initiated in very young mice is a delay in disease onset (Zhu et al., 2002), while administration closer to disease onset results in a slowing of disease progression after onset (Kriz et al., 2002). Regardless, both treatment paradigms provided a comparable extension of survival (Kriz et al., 2002). Since minocycline also exerts an antiapoptotic property to neurons (Zhu et al., 2002), it is unclear in which cells the minocycline effect was active. Either way, supported by its beneficial effect in ALS mice, minocycline has been proposed for clinical trial (Gordon et al., 2004 and Pontieri et al., 2005).

Cyclooxygenase-2 (COX-2), produced in abundance by microglia and other inflammatory cells (but also by neurons and astrocytes), plays a key role in stimulating production of proinflammatory cytokines, and COX-2 expression is induced in spinal cords of ALS patients (Yasojima et al., 2001 and Yiangou et al., 2006).

In mice, use of a COX-2 inhibitor (celecoxib) prolonged survival by slowing disease onset (Drachman et al., 2002) but disappointingly did not alter progression after onset. Subsequent test of celecoxib in a human trial failed to provide a benefit in ALS (Cudkowicz et al., 2006).

Additional evidence implicating microglia in pathogenesis of ALS arose from forcing activation of the immune system using chronic administration of lipopolysaccharide (LPS), a well-known microglial activator. Such treatment exacerbated SOD1 mutant-mediated disease in mice (Nguyen et al., 2004).

Similarly, inducing microglial activation by deleting the fractalkine receptor, a cytokine receptor expressed by microglia, mildly accelerated neuronal loss in mutant SOD1 mice (Cardona et al., 2006). Increased levels of several proinflammatory factors have been described in the spinal cords of mutant SOD1 mice even before motor neuron loss (Elliott, 2001, Hensley et al., 2003 and Yoshihara et al., 2002).

Cerebrospinal fluid (CSF) and serum of ALS patients also contain increased levels of inflammation related factors, including complement proteins and monocyte chemoattractant protein-1 (MCP-1; Goldknopf et al., 2006 and Simpson et al., 2004).

One particularly interesting cytokine upregulated in mutant SOD1 mouse spinal cords which could play a role in motor neuron degeneration is tumor necrosis factor- (TNF). TNF has been shown to be produced in higher levels by adult hSOD1G93A microglial cells when stimulated with LPS compared to nontransgenic microglial cells (Weydt et al., 2004). Administration of a TNF antagonist yielded a mild increase in survival in mutant SOD1 mice (West et al., 2004). A direct toxic effect of activated microglial cells on motor neurons has been shown in a coculture system and seemed to implicate nitric oxide (Zhao et al., 2004). Microglial cells could therefore be an important player in a Fas-ligand (FasL)-induced apoptosis pathway within motor neurons that is driven by nitric oxide synthesis (discussed in more detail below). In addition, mutant SOD1, which has now been reported to be released by motor neurons, is a potent activator of microglial cells (Urushitani et al., 2006), emphasizing the likely crosstalk between motor neurons, microglial cells, and potentially other nonneuronal cells that may cooperate to drive disease progression.

An Outside-In Cascade Triggering a Cell-Death Pathway Intrinsic to the Motor Neuron

At least a partial explanation for the selectivity of motor neurons to SOD1 mutant toxicity has arisen from identification of a motor-neuron-specific cell-death pathway. Not surprisingly, embryonic motor neurons extracted from spinal cords of mutant SOD1 mice are more susceptible to toxic insults (Kruman et al., 1999, Raoul et al., 2002, Spalloni et al., 2004b and Van Den Bosch et al., 2004) and have modified electrophysiology and altered excitability (Pieri et al., 2003 and Spalloni et al., 2004a) compared to neurons from normal mice or mice expressing high levels of hSOD1WT protein. These neurons also have what has been proposed as a motor-neuron-specific cell-death pathway downstream of the Fas death receptor. Increased susceptibility to death triggered by the Fas receptor requires nitric oxide, the Fas death domain-associated protein Daxx, and p38 kinase activation, which subsequently drives nitric oxide production, thus generating a feed-forward amplification loop (Raoul et al., 1999, Raoul et al., 2002 and Raoul et al., 2006). Cultured motor neurons from transgenic mutant SOD1 mice have an increased susceptibility to activation of this pathway, which is apparently also activated in presymptomatic ALS mice (Raoul et al., 2002, Raoul et al., 2006 and Wengenack et al., 2004). Given the acceleration of disease by mutant microglia, it is plausible that SOD1 mutant-induced release of nitric oxide by microglia drives this death cascade within motor neurons.

Astrocytes in ALS: Gatekeepers of Synaptic Glutamate and Excitoxicity

Astrocytes are essential partners of motor neurons, providing them with trophic support and mediating rapid recovery of synaptic glutamate through the action of the glial glutamate transporter EAAT2 (also referred to as GLT-1 in rodents).

Astrocytes insert finger-like processes loaded with this transporter into and surrounding synapses, thereby facilitating rapid removal of synaptic glutamate following its release from the presynaptic terminal (as detailed in Figure 3).

Glutamate-mediated excitotoxicity can induce neuronal damage and death as a consequence of repetitive firing, which in turn drives calcium entry through calcium-permeable AMPA glutamate receptors.

Within the spinal cord, preventing glutamate-driven excitotoxicity is primarily the job of astrocytes. Motor neurons have an inherently high sensitivity to excitotoxicity since their glutamate receptors have a lower proportion of the GluR2 subunit (Van Damme et al., 2002) relative to many other types of neurons. Regulation of the Ca2+ permeability properties of GluR2 is due to posttranscriptional RNA editing, with only AMPA receptors containing edited GluR2 resistant to Ca2+ entry. Already with low GluR2 levels, the efficiency of mRNA editing is reduced in the spinal cord motor neurons of ALS patients compared to controls (Kawahara et al., 2004 and Takuma et al., 1999). Altered efficiency of editing was not found in hSOD1G93A or hSOD1H46R rats (Kawahara et al., 2006), thereby demonstrating that change in mRNA editing is not required for familial disease.

Astrocytes also respond to damage in many settings by what is referred to as activation. This includes an increase in assembly of their intermediate filaments (assembled from glial fibrillary acidic protein, GFAP) and an increase in the number and size of processes extended from the cell body (Figure 3). Astrocyte activation is seen in spinal cords of ALS patients and SOD1 mutant mice (Hall et al., 1998, Levine et al., 1999 and Schiffer et al., 1996). Indeed, for mice expressing the dismutase inactive mutant hSOD1G85R, a very prominent early pathology that increases markedly during disease course is the presence of SOD1-containing inclusions within activated astrocytes (Bruijn et al., 1997).

Altered glutamate handling is one of the few firm mechanistic links between sporadic and SOD1 mutant-caused ALS. Diminished glutamate transport has been reported in synaptosomes obtained from affected CNS tissues of ALS patients (Rothstein et al., 1992), and levels of EAAT2 are reduced in the motor cortex and spinal cord of ALS patients (Fray et al., 1998, Maragakis et al., 2004, Rothstein et al., 1995 and Sasaki et al., 2000), spinal cords of mutant SOD1 mice (Bruijn et al., 1997) and rats (Howland et al., 2002), as well as in a model of Sindbis virus-induced motor neuron disease (Darman et al., 2004). This is functionally of consequence for familial ALS in that hSOD1G93A mice heterozygous for EAAT2 develop earlier-onset disease (Pardo et al., 2006), while drugs that increase EAAT2 activity extend survival (Ganel et al., 2006 and Rothstein et al., 2005). Indeed, screening of FDA-approved drugs for those that could elevate EAAT2 activity has identified a CNS-penetrating β-lactam antibiotic, ceftriaxone, as a transcriptional inducer that modestly extends survival in hSOD1G93A mice (Rothstein et al., 2005). An inhibitory effect on quantal glutamate release is also believed to underlie the mild slowing of disease course in human ALS from the only FDA-approved treatment (riluzole; Doble, 1996).

Additional implication for an astrocytic contribution in ALS is demonstration in cell culture that activated astrocytes induce embryonic motor neuron degeneration via their production of NGF, which in the presence of low concentrations of nitric oxide can induce apoptosis of motor neurons through the p75NTR receptor. Indeed, relative to resting astrocytes, activated astrocytes in the spinal cord of hSOD1G93A mice have increased NGF synthesis (Pehar et al., 2004).

Muscle Involvement in ALS: Targets for Therapy?
Among the earliest events in the human ALS- and SOD1-mediated disease is withdrawal of the motor axons from the neuromuscular synapse (Fischer et al., 2004, Frey et al., 2000, Pun et al., 2006 and Schaefer et al., 2005). This denervation generates the progressive paralysis of ALS, a consequence of which is muscle atrophy (Cifuentes-Diaz et al., 2001). Mutant SOD1 is also expressed by muscle, but whether its presence there contributes to pathogenesis is not yet established.

A test of whether enhanced muscle mass and strength per se can be beneficial in ALS has been attempted by repetitive injection into hSOD1G93A mice of an antibody to myostatin, a secreted protein whose action inhibits muscle growth. Initiated before disease onset, this produced enhanced muscle mass initially, but the effect on limb muscles was short lived, yielding neither a survival benefit nor preservation of muscle mass throughout disease progression (Holzbaur et al., 2006).

On the other hand, muscle hypertrophy induced by agents such as insulin-like growth factor-1 (IGF-1) or growth hormone (Dobrowolny et al., 2005, Kaspar et al., 2003 and Kaspar et al., 2005) has led to significant life extensions in ALS transgenic mice. For the IGF-1 studies, not only was there muscle hypertrophy but also concomitant stimulation of muscle satellite cell proliferation and an increase of centrally nucleated muscle fibers, indicating regeneration. Although in one instance IGF-1 synthesis was mediated by a transgene expressed only by the muscle (Dobrowolny et al., 2005), it has not been established if the secreted IGF-1 acts on the muscle, the motor neuron, or both. At a minimum, these efforts illustrate the possibility of using trophic factor synthesis by muscle (see section below) as a way of rescuing motor neurons. Added to this, a number of studies have shown that exercise is beneficial in ALS transgenic animals (Kirkinezos et al., 2003 and Veldink et al., 2003), with exercise and IGF-1 exhibiting a synergistic effect resulting in an increase in median life span by 83 days (Kaspar et al., 2005). The mechanism underlying this synergism is not known, and the relevance for human ALS unclear, as no consensus has emerged from several efforts to evaluate a benefit of exercise in human disease (Drory et al., 2001, Pinto et al., 1999 and Scarmeas et al., 2002).

A Model for Selective Vulnerability of Motor Neurons in ALS
Combining the lessons from multiple animal models used a determine pathogenic mechanisms in inherited ALS, the central insight is that although motor neuron degeneration and death is the primary cause of the progressive paralysis, toxicity is produced by damage developed not only within motor neurons, but also by other nonneuronal neighbors (Figure 3).

Onset and Early Phase
Damage directly within motor neurons is an important determinant of disease onset. The earliest event, prior to disease symptoms, is retraction of motor axons from their synapses onto muscles. During this phase, mutant SOD1 primarily acts directly within motor neurons, with aggregation of misfolded SOD1 damaging cellular machinery, especially mitochondria, so as to inhibit one or more normal functions. Anterograde axonal transport is also inhibited, as the consequence either of mitochondrial damage or of disorganization of axonal neurofilaments. SOD1 mutant action within the motor neuron is amplified by action within other cell types, especially the microglia which respond to the initial damage by amplifying it, thereby driving more rapid disease progression and spread.

Symptomatic
The symptomatic phase is characterized by a massive activation of microglia and astrocytes, in addition to continuing damage within motor neurons themselves. Misfolded SOD1 mutant within astrocytes, as well as their activation in response to neuronal damage, induces loss of the EAAT2 glutamate transporter, reducing rapid recovery of synaptic glutamate and driving an excitotoxic response that includes excessive influx of Ca2+ during repetitive firing of glutamate receptors on motor neurons. SOD1 mutant action directly within microglial and astrocytic cells may provoke reduced secretion of trophic molecules or increased release of toxic factors (e.g., nitric oxide, TNF, or mutant SOD1 secreted or released by cell leakage or lysis). The latter molecules drive spread of toxicity to neighboring cells. Mutant SOD1 can itself activate microglia. In return, microglial cells, through secretion of toxic factors, harm their environment, including motor neurons.

End Stage
At end stage, denervation has produced paralysis accompanied by muscle atrophy. Activated microglia or astrocytes produce diffusible toxic products, including nitric oxide and/or TNF, which accelerate disease spread. Loss of EAAT2 glutamate transporters from astrocytes drives repetitive firing of glutamate receptors and an ensuing excitotoxicity that is accompanied by excessive Ca2+ entry that further damages mitochondria and finally triggers a caspase-dependent cell suicide pathway within motor neurons.

Thus, selective sensitivity of motor neurons to ubiquitously expressed SOD1 mutants derives from the combination of risk factors shared by those neurons and their most intimately associated cellular neighbors. Seen this way, it is clear that the question is not whether the mitochondrial damage and ensuing oxidative damage hypothesis is right, any more than it is whether the hypotheses for glutamate-linked excitotoxicity, axonal transport clogging, loss of trophic support, or excessive release of toxic species are right. All these mechanisms (which have also been invoked in sporadic ALS) are right and all combine to underlie selective vulnerability. This is very good news for the design of therapies: intervention at any of these steps can, in principle, disrupt the toxic cascade between the motor neuron and its neighbors.

Growth Factor Therapies in ALS
Use of trophic factors has dominated recent efforts in clinical trials in ALS. All of these efforts were predicated on the hope that whatever the primary disease provoking insult, provision of increased levels of factors that could be trophic for motor neurons would be of benefit. These have lead to nearly uniform disappointment, providing little or no benefit for IGF-1 (Borasio et al., 1998 and Lai et al., 1997) and no benefit for BDNF (BDNF Study Group, 1999) or CNTF (ALS CNTF Treatment Study Group, 1996). Aside from the absence of evidence that growth factors were limiting in ALS, the weakness of these approaches was insuring adequate delivery to the right cells. One method for eliminating blockage of drug delivery by the blood-brain barrier is direct, continuous infusion into either the brain (ICV) or spinal cord (intrathecal; Figure 4B). Continuous intrathecal administration of BDNF was found to be without benefit, but ICV infusion of IGF-1 extended survival in mice (Nagano et al., 2005a), and a phase I clinical trial using the same strategy showed a modest benefit to patients without adverse effects (Nagano et al., 2005b).

Full-size image (95K)

Figure 4. New Directions for Therapies in ALS Using Viral Delivery or Direct Infusion into the Brain or Spinal Cord

(A) Delivery by injection into muscle of viral particles (AAV) encoding trophic factors such as IGF-1 or VEGF. Viral particles are taken up at the synapse by the motor neuron terminal and retrogradely transported to the motor neuron cell body in the spinal cord. After translocation into the nucleus, the episomal viral genome is stably expressed for at least 1 year, converting motor neurons into factories for secretion of trophic factors. Trophic factors could also be synthesized directly in the muscle, increasing their support of the motor neuron terminal. This approach is also suitable for viral delivery of transcription-based siRNA approaches in which mutant SOD1 mRNAs (or specific cellular mRNAs) are targeted for degradation selectively within motor neurons. An alternative is for direct injection of viral particles into the spinal cord. This approach is more invasive but has the advantage of targeting not only the motor neurons but also nonneuronal neighboring cells of the motor neurons.

(B) Direct infusion into the spinal cord (intrathecal) or brain (intracerebroventricular, ICV) of trophic factors directly into the CSF yields widespread delivery throughout the brain and spinal cord. Alternatively, ICV infusion of antisense oligonucleotides can reduce mutant SOD1 by targeting intranuclear degradation of its mRNA both within the motor neurons and the surrounding microglial and astrocytic cells.

In animal models, efficacy in growth factor delivery has been achieved by either of two approaches: viral-mediated gene delivery to convert target cells into factories that secrete trophic factors and infusion of trophic molecules directly into the brain or spinal cord (Figure 4). For the first approach (Figure 4A), an IGF-1-encoding gene was inserted into adenoassociated virus (AAV), a small replication-defective virus that is episomally maintained within the nucleus of a host cell. Injection into muscle yielded viral expression within motor neurons for at least 1 year. Use of this strategy slowed disease progression in hSOD1G93A mice even when initiated after disease onset (Kaspar et al., 2003). Not surprisingly, success required the right trophic factor. The same AAV producing GDNF used in the same manner provided only a very modest benefit (Kaspar et al., 2003), consistent with prior efforts using GDNF encoded by adenovirus (Acsadi et al., 2002) or AAV (Wang et al., 2002) delivered intramuscularly.

With the potential that variants in the VEGF gene contribute to some examples of ALS (Table 1), delivery of an integrating lentivirus encoding VEGF (and pseudocoated so as to be retrogradely transported) extended survival of hSOD1G93A mice (Azzouz et al., 2004). So too did continuous ICV infusion of recombinant VEGF protein into the CSF (Figure 4B): disease onset was delayed and survival extended by 22 days (Storkebaum et al., 2005). This ICV delivery of VEGF was especially effective in slowing forelimb paralysis, suggestive of a higher concentration of VEGF closer to the site of infusion. A modest benefit was seen even when VEGF treatment was initiated after symptomatic onset. Unresolved is which cells are targeted by this VEGF. Weekly intraperitoneal injection of VEGF also has been reported to slow disease in hSOD1G93A mice (Zheng et al., 2004 C. Zheng, I. Nennesmo, B. Fadeel and J.I. Henter, Vascular endothelial growth factor prolongs survival in a transgenic mouse model of ALS, Ann. Neurol. 56 (2004), pp. 564–567. Full Text via CrossRef | View Record in Scopus | Cited By in Scopus (35)Zheng et al., 2004), raising the possibility that important target cells are outside the CNS, including those in the vasculature.