Protein to protein interaction, seeded aggregation and propagation in ALS

Familial ALS is classified into two groups, with and without aggregation of TDP-43. The cytoplasmic mislocalisation and aggregation of TDP-43 is now the major pathological hallmark for sporadic ALS. Similar TDP-43 pathology is found in familial ALS patients with mutation of optineurin, C9ORF72 and VCP, but is not observed in those with mutation of SOD1 and FUS. In the past 2 years, there have been cumulative data that normal and mutant SOD1 and normal TDP-43 can undergo a similar seeded aggregation,6 although the pathophysiology of TDP-43 and mutant SOD1 is different and difficult to discuss in the same platform.

TDP-43 is a nuclear protein and consists of two RNA recognition motifs (RRM1 and RRM2) and a proteolytic C terminal domain, where most of the mutations are found. C terminal fragments of TDP-43 have been identified in the cytoplasmic inclusions of sporadic ALS patients.2 3 TDP-43 readily aggregate in vitro which is enhanced by ALS associated mutation,7 and C terminal fragments may play a critical role in TDP-43 aggregation.8 9 In cultured cells, misfolded forms of TDP-43 can initiate intracellular aggregation and induce the misfolding and subsequent aggregation of the respective native proteins.8 9 What triggers TDP-43 aggregation has not been determined but increased protein concentration due to disruption of microtubule transport9 or chronic cellular stress and defects in stress granule disassembly10 occurring with the ageing process may converse to its initiation of this pathological event.

SOD1 is a stable dimer and is highly resistant to proteolytic degradation. More than 150 point mutations have been identified in SOD1, occurring at almost every position; these might impair the stability of the protein and influence the formation of aggregates and inclusion bodies. Both native and mutant forms of SOD1 associated with each other, and some types of mutant SOD1, easily form aggregates and fibrils under denaturing conditions.5 6 11 Importantly, removal of the misfolded seeds does not stop aggregation of endogenous SOD1, indicating that the newly formed aggregates can act as templates for the subsequent misfolding of additional native SOD1.11

Neuron to neuron propagation in ALS

Cellular proteins could be released from neurons via vesicle mediated exocytosis or direct leakage through a damaged cell membrane. Exosomes are small lipid membranous microvesicles which are usually formed by the fusion of multivesicular bodies with the plasma membrane. The biologically important functions of exosomal release are the secretion of proteins in order to remove unwanted proteins, the delivery of signals to other cells and the transfer of pathogens among cells. Exosomes derived from prion infected neuronal cell lines are efficient initiators of prion propagation in uninfected recipient cells and non-neuronal cells. Indeed, mutant SOD1 has been proposed to be actively secreted with neurosecretory vesicles through an abnormal interaction with chromogranins A and B.12 Protein aggregates extracted from mutant SOD1 expressed cells could penetrate into cells by macropinocytosis.6 Importantly, a passive immunisation approach based on intracerebroventricular infusion in G93A-SOD1 mice of monoclonal antibodies specific to misfolded forms of SOD1 succeeded in reducing the level of mutant SOD1 in the spinal cord and in prolonging the lifespan of G93A-SOD1 mice to a minor degree.13 The exact mechanisms by which anti-SOD1 antibodies confer protection are not yet fully understood but the toxic effect of the excreted mutant SOD1 might be blocked by the antibodies.

Glia to neuron propagation of toxicity in ALS

Toxic factors secreted or released from microglia and astrocytes damage motor neurons in ALS mouse models of mutant SOD1, which may lead to disease progression.14 15 In vivo transplanted glial restricted precursor cells expressing mutant human SOD1 (G93A) survive and differentiate efficiently into astrocytes in the spinal cords of mice, and graft derived mutant human SOD1 astrocytes induce ubiquitination and death in host motor neurons. Only mutant human SOD1 expressing astrocytes can induce wild-type motor neuron death in mice.16 Conversely, siRNA knockdown of SOD1 in astrocytes significantly attenuates toxicity against motor neurons. In ALS, astrocytes might induce and regulate unknown toxic factors that have a non-cell autonomous effect on motor neurons when released.17

Focality of clinical onset and possible influencing factors

The clinical manifestations of ALS appear to start at a focal body region and then spread to other regions over time.18–22 The focality of clinical onset is an important issue to discuss because there must be an onset mechanism different from progression pathology, whether or not cell to cell propagation underlies the progression of ALS. At the clinical onset, almost equal proportions of patients have bulbar region, upper limb or lower limb symptoms,19 20 with bulbar cases slightly less frequent (21–25% of all patients).18 21 22 Respiratory symptoms rarely occur first. Thus it seems that the disease can begin stochastically at any region. However, in patients with unilateral upper limb onset, a different frequency of onset between the right and left sides has been reported. Ravits et al19 reported that 34 of 100 ALS patients had onset in the upper limbs: 24 (70.6%) in the right and 10 (29.4%) in the left limb. In 53 ALS patients with unilateral upper limb onset reported by Körner et al,20 onset was on the right in 31 (58.5%) and on the left in 22 (41.5%). This difference between the right and left sides may relate to handedness. Turner et al23 demonstrated that, for patients with upper limb onset, there was statistically significant concordance between side of onset and handedness: 97 of 151 (64%) showed concordance overall, with 90 of 139 right-handed patients having right-sided onset and seven of 12 left-handed patients having left-sided onset. In contrast, no predominance for side or footedness has been found in patients with lower limb onset. These findings suggest that frequent use may influence symptom onset because the dominant hand will be used more frequently and both legs will be almost equally used in daily activity although a case control study revealed no association between physical activity and the risk of developing ALS.24 The corticomotoneuronal organisation in the dominant and non-dominant hand could also explain the difference in right and left onset.

Different motor neurons may also have different vulnerabilities. In a three-dimensional study of the distribution of anterior horn cells in the cervical eighth segment of the human spinal cord obtained at autopsy from patients with ALS, zones of focal and patchy loss of motor neurons were identified without consistent severity or distribution at different sectional levels.25 According to the investigators, these findings imply that different pools of motor neurons have varying resistance to disease. In limb onset cases, the initial symptoms present most commonly in distal muscles. In an analysis of 1200 patients, Haverkamp et al22 found that 73% of non-bulbar patients first exhibited symptoms in the distal portions of the limbs. In 24 autopsies of ALS patients, Tsukagoshi et al26 found more marked anterior horn cell loss at the eighth cervical segment, which mainly innervates the distal hand muscles, than at the sixth cervical segment innervating more proximal upper limb muscles. The distal predominance in the initial symptoms does not necessarily reflect only axon length dependent vulnerability because bulbar and lower limb onset occur at the same frequency. By using quantitative concentric needle EMG in the early stages of ALS, de Carvalho et al27 showed that similar abnormalities in motor unit potentials (MUPs) were found in paraspinal and limb muscles in the same spinal segment, both in the sixth cervical segment and the fifth lumbar segment, although fasciculation potentials were more frequent in limb muscles than in paraspinal muscles, and fibrillations were most frequent in the distal tibialis anterior muscle. They concluded that the results are consistent with generalised involvement of motor neurons in the spinal segments in the early stages of ALS progression. The vulnerability might differ not only in motor neurons themselves but also in the segments or sites within a segment.

Another vulnerability may depend on axonal excitability. In ALS, muscle wasting preferentially affects the thenar complex, including the abductor pollicis brevis and first dorsal interosseous (FDI) muscles, compared with the hypothenar muscle (abductor digiti minimi (ADM)). This peculiar pattern of dissociated atrophy of the intrinsic hand muscles, termed the ‘split hand’, is nearly specific in ALS.28 Using threshold tracking, Bae et al29 investigated the different membrane properties of FDI and ADM motor axons in the same ulnar nerve at the same site in 12 normal subjects, and demonstrated that nodal persistent sodium conductances are more prominent in FDI axons than in ADM axons, and therefore excitability is physiologically higher in FDI axons. The higher excitability in FDI axons than in ADM axons may explain the preferential involvement of FDI over ADM in ALS.

There can also be selective vulnerability by physiological subtypes of motoneurons. Pun et al30 showed in two mouse models of motoneuron disease (G93A SOD1 and G85R SOD1) that axons of fast fatigable motoneurons were affected synchronously, long before symptoms appeared, and fast fatigue resistant motoneuron axons were affected at symptom onset, whereas axons of slow motoneurons were resistant. In later stages of human sporadic ALS, motor units with high recruitment thresholds for voluntary contraction, which correspond to the larger motoneurons, were more affected in patients with more severely affected FDI muscles.31

Spreading patterns of clinical manifestations

In the progression of ALS, muscle weakness regionally spreads over time, and most UMNs and LMNs ultimately degenerate. Spread of clinical manifestation, as well as exacerbation of focal motor symptom, should reflect an increase in the number of affected UMNs and LMNs although disturbance in a given motoneuron can have widespread effects because of synaptic dysfunction and loss of control at a distance in large and complex neural networks. If cell to cell propagation of disease property would make healthy adjacent motoneurons ill, the simplest pattern of symptom spread should be the contiguous spread into adjacent body regions. However, although spread is contiguous in many ALS patients, it is sometimes non-contiguous.

Ravits et al19 32 described, from cross sectional clinical observations, that the focal and discrete motor manifestations of ALS present at clinical onset spread outward into contiguous body regions. In 14 of 19 ALS nervous systems examined at autopsy, the regional lower neuron loss within the nervous system was graded radially away from the region of onset, suggesting contiguous spreading of the degenerative process.33 This contiguous spreading pattern has also been found by other investigators who observed cohorts of ALS patients longitudinally.20 21 Horizontal spreading along the neuroaxis occurs more frequently than radial spreading—that is, the contralateral homologous limb tends to be involved sooner than other regions in patients with unilateral limb onset18 20 34 although such a directionality was not observed in one study.35 In patients with bulbar onset, radial spreading to the upper limbs occurred sooner than spreading to the lower limbs in one study21 although the difference was not statistically significant in another study.18 These observations suggest that one factor influencing the time to symptomatic spread could be distance from the onset site. Another important feature of radial spreading is the preferential involvement of the side ipsilateral to the side of onset—that is, the leg ipsilateral to the onset arm and the arm ipsilateral to the onset leg tend to be involved before the contralateral limbs.18 20 34 The underlying mechanisms of the preferential ipsilateral spread have not yet been elucidated but radial propagation on the ipsilateral side or the neural connections projecting to the ipsilateral cervical and lumbar spinal motoneurons, such as the corticospinal system, might be involved.

Spread to non-contiguous regions has also been documented although less frequently than contiguous spreading. Fujimura-Kiyono et al21 showed that, among 150 patients with sporadic ALS who underwent follow-up every 3 months, the second region involved was a distant (non-contiguous) region in 14% of 52 patients with lower limb onset and in 29% of 32 patients with bulbar onset. A non-contiguous spread pattern was also found in 14.8% of 282 sporadic ALS patients investigated retrospectively by Gargiulo-Monachelli et al35 If ALS is a disease whose pathological process starts as a focal ‘seed’ and then propagates outward, the non-contiguous spreading pattern implies that some cases start in multiple foci. This possibility is supported by the existence of patients with clinical onset in two regions at almost the same time. Such cases were found in 10.6% of a cohort of 150 patients with sporadic ALS.21 Körner et al20 also found patients with lower limb onset who already had bulbar involvement at disease onset. Among 19 autopsy cases of sporadic ALS, the loss of motoneurons was greatest in a region different than that identified as the region of clinical onset in two cases.33 If only one seed is involved, one possible explanation for non-contiguous spreading is anterograde trans-synaptic propagation. Supporting this, the spread between bulbar and spinal regions occurs preferentially in the rostral–caudal direction than caudal–rostral direction18–21 although this difference was not obvious in one study.35 Assessments of the cumulative occurrence of symptomatic involvement, based on information from serial questionnaires, demonstrated that limb symptom accrual in 155 bulbar onset ALS patients was significantly faster than bulbar symptom accrual in 387 ALS patients with unilateral limb onset over 6 years of follow-up.18 On Kaplan–Meier plots, duration from bulbar onset to limb involvement was shorter than the reverse (9 months from bulbar regions to upper limb vs 17 months from upper limb to bulbar regions and 14 months from bulbar regions to lower limb vs 27 months from lower limb to bulbar regions until occurrence in 50% of patients).21 In the monkey, it was demonstrated that some pyramidal tract neurons in the hand and trunk areas of the primary motor cortex have multiple axon branches projecting to motor neuron pools in non-adjacent spinal segments.36

However, it is important to note that the degenerative process advances insidiously, and at least one-third of LMNs for a given muscle were estimated to be lost before the clinical signs manifest. Indeed, widespread abnormalities on needle electromyography are often found even if clinical deficit is localised, indicating that the clinical symptom manifests according to the balance between the rate of motoneuron loss and the ability of collateral sprouting of the remaining motoneurons which compensates the loss. Therefore, intensive needle electromyographic studies are needed for the analysis of non-contiguous spreading.

Relation between upper and lower motor neuron involvement

In ALS, neurodegeneration occurs in both the UMNs in the cortex and the LMNs in the brainstem and spinal cord, and a variable combination of UMN and LMN involvement produces a wide variety of motor symptoms. However, the relationship between UMN and LMN involvement is not yet fully clear. There is ongoing discussion as to whether the UMNs or LMNs are primarily involved (anterograde ‘dying forward’ transneuronal degeneration vs retrograde ‘dying back’ degeneration) or each is independently affected.

With regard to the onset site, Ravits et al32 reported that during the early stages of ALS, UMN and LMN signs were maximal in the same peripheral body region. They speculated that the triggering pathogenic event may be distributed by transneuronal signalling or axonal transport, which is relevant to the trigger and initial distribution between UMNs and LMNs. In contrast, Körner et al20 did not find a correspondence of onset regions between UMNs and LMNs in their cohort. They found that the frequency of UMN signs was independent of the onset region and was highest in the proximal legs at any given time although the side of involvement corresponded with the UMN and LMN signs. They concluded that the UMNs of the lumbar region are more vulnerable.

With regard to regional spread, Ravits et al found that the outward spread of UMN and LMN signs differed during disease progression; these signs either spread to different peripheral body regions or to the same peripheral body region but to different extents, and there was no correlation in the severity of UMN and LMN involvement. Other neuropathological and neurophysiological findings seem to be consistent with the independent involvement of UMNs and LMNs. In 12 autopsy cases of the classical sporadic form of ALS, the pyramidal cells in layer 5 of the tongue and foot areas of the precentral gyri were significantly smaller in patients than in controls but no correlation was found between the mean sizes of pyramidal cells and the numbers of surviving LMNs in the hypoglossal nucleus or in the anterior horn of the fourth lumbar segment of the spinal cord.37 In 18 cases of sporadic motor neuron disease, there was poor and negative correlation between the densities of the spinal motor neurons in the 6th–8th cervical segments and of the corticomotoneurons in the corresponding areas.38 In a longitudinal neurophysiological study, Attarian et al39 investigated the changes over time of MUPs in the extensor carpi radialis and the same single motor unit responses induced by transcranial magnetic stimulation (TMS). They found no correlation between UMN and LMN dysfunction in the course of disease progression. A positive correlation was observed between macro-MUP size and the amplitude of the excitatory responses to TMS only in the earliest stage of the disease but this correlation disappeared within the second year after onset. Analysis of MUPs in the sternomastoid, which receives both contralateral and ipsilateral corticospinal innervation, and in the trapezius, which usually only receives contralateral innervation, revealed symmetric changes in both muscles, and the two muscles were equally affected, suggesting that UMNs and LMNs are independently involved, given the characteristic asymmetry of ALS.40

In contrast, Körner et al20 observed that UMN and LMN signs predominantly spread together, rather than independently, as Ravits et al described. In the Körner et al study, LMN involvement was more distinctive than UMN involvement at disease onset, which may support the hypothesis of initial involvement of LMNs with a retrograde ‘dying back’ process leading to secondary UMN degeneration.

On the other hand, the findings from a study using TMS in patients with familial ALS with SOD1 mutations could support the hypothesis of initial involvement of UMNs with an anterograde ‘dying forward’ transneuronal process leading to secondary LMN degeneration. Cortical hyperexcitability, which has been shown in the early stage of sporadic ALS,41 42 was also found in three presymptomatic SOD1 mutation carriers, and subsequently these three individuals developed the clinical features of ALS.43 Loss of inhibitory cortical interneurons is assigned as one possible mechanism for such cortical hyperexcitability in the early stage of the disease. These cells may be the earliest to be involved. The increased concentration of extracellular glutamate may also cause the hyperexcitability of corticospinal neurons.

Interestingly, the split hand in ALS may have a cortical basis.44 A study using TMS demonstrated that the ratios of motor evoked potentials to compound muscle action potential were significantly larger for the thenar complex in normal subjects. This suggests there is a stronger corticomotoneuronal input to these muscles, and relatively extensive corticomotoneuronal inputs on the thenar spinal anterior horn cells would be associated with more glutamate release, which may cause preferential degeneration in ALS through glutamate induced excitotoxicity. A similar rationale might exist for the preferential involvement of the dominant hand in ALS. A TMS study in relation to handedness in normal subjects demonstrated a shorter cortical silent period, reflecting reduced inhibitory function, in the dominant hand.45 The corticomotoneuronal organisation might be relevant to the vulnerability of LMNs rather than to the propagation of pathogenic molecular events.