SPG4 Hereditary Spastic Paraplegia: From Etiology to Therapy

Abstract

Hereditary spastic paraplegias (HSPs) comprise a heterogeneous group of heritable neurodegenerative disorders resulting from mutations in a wide variety of genes. HSP locomotor symptoms include lower limb weakness and spasticity that arise from progressive degeneration of corticospinal axons projecting from the motor cortex to the distal spinal cord. Ensuing gait defects typically lead to wheelchair dependence. Degeneration of other tracts can also occur, expanding the constellation of symptoms to potentially include urinary, fecal, and speech problems, as well as intellectual disability. SPG4‐HSP, the most common variant, is caused by mutations in the SPAST gene, which encodes spastin, a microtubule‐severing protein with membrane‐related properties. Significant progress has been made in developing preclinical models for SPG4‐HSP and in elucidating its mechanistic etiology. On this basis, progress is being made on developing a flexible therapeutic regimen for patients at all stages of disease progression. © 2026 The Author(s). Movement Disorders published by Wiley Periodicals LLC on behalf of International Parkinson and Movement Disorder Society.

Hereditary spastic paraplegias (HSPs) are rare diseases in which degeneration of the corticospinal nerve tracts leads to lower limb spasticity and gait deficiencies. ¹ Depending on the gene mutated and even within a family with the identical gene mutation, symptoms can vary a great deal in their severity, age of onset, and the degree to which other nerve tracts degenerate in addition to the corticospinal tracts. Symptoms can also include problems with upper limbs, speech, cognition, sexual function, bladder control, and fecal control. ² , ³ , ⁴ , ⁵ , ⁶ , ⁷ , ⁸ The number of genes whose mutations give rise to HSPs is growing, with over 90 currently identified. ⁹ Many cases are misdiagnosed and mistaken for other diseases, such as cerebral palsy, suggesting that HSPs are not as rare as once believed.

By far the most common form of HSP in which the gene is known, termed HSP‐4 or SPG4‐HSP, is an autosomal dominant disease resulting from mutations of the SPAST gene. ¹⁰ , ¹¹ For a variety of reasons, SPG4‐HSP is especially challenging in terms of understanding its mechanistic basis and developing therapies accordingly. Even today, there are no treatments for patients beyond physical therapy to maintain muscle tone and Botox injections and baclofen (administered orally or via intrathecal pump) to reduce spasticity. ¹² , ¹³ , ¹⁴

The SPAST gene encodes a protein called spastin. SPAST has two start codons, and therefore the same messenger RNA (mRNA) encodes both a longer isoform of spastin and a shorter isoform (Fig. 1A). ¹⁵ , ¹⁶ These two isoforms are quite different in their properties and functions, suggesting that they might act as two entirely different proteins (that might still functionally interact). The first start codon is a weak one, resulting in low expression of the longer isoform of spastin called M1, whereas the second start codon is strong, resulting in high expression of the shorter isoform called M87 (Fig. 1A). The M1‐specific region, consisting of 86 amino acids, includes a hydrophobic portion that likely makes it prone to misfolding, which is presumably why the weak first start codon evolved to ensure extremely low expression. Targeted degradation of M1 also contributes to its low abundance. ¹⁷ , ¹⁸ , ¹⁹ M1 is detectable, albeit not easily, in motor cortex and spinal cord of adult rodents. ²⁰ , ²¹ , ²² , ²³

FIG. 1.

Overview of SPAST gene structure and spastin isoform functions. (A) Schematic representation of the SPAST gene and encoded spastin proteins. The top panel shows the chromosomal location, exons (filled boxes), and introns (gray lines) of the gene. The bottom panel depicts the spastin proteins with functional domains highlighted, including the N‐terminal region and hydrophobic domain (HR), microtubule‐interacting and trafficking (MIT) domain, microtubule‐binding domain (MTBD), and AAA ATPase domain. Arrows indicate the translational start sites of the M1 and M87 isoforms. (B) Illustration of the M87 isoform of spastin promoting microtubule severing in an ATPase‐dependent manner, highlighting its role in microtubule remodeling. (C) Illustration of the M1 isoform of spastin integrating into the endoplasmic reticulum membrane, emphasizing its unique N‐terminal domain responsible for membrane association and potential impact on the endoplasmic reticulum morphology. [Color figure can be viewed at wileyonlinelibrary.com]

M87 is a well‐characterized microtubule‐severing enzyme that forms hexamers on the surface of the microtubule to generate forces needed to break one microtubule into two (Fig. 1B). ²⁴ , ²⁵ Similar to another microtubule‐severing protein called katanin, ²⁶ , ²⁷ , ²⁸ M87 cuts microtubules free of the centrosome and breaks microtubules into shorter pieces. ²⁵ , ²⁹ , ³⁰ , ³¹ , ³² , ³³ , ³⁴ , ³⁵ , ³⁶ , ³⁷ , ³⁸ , ³⁹ , ⁴⁰ This is important because unanchored short microtubules are more readily transported and reconfigured to accommodate shape changes in the life of the neuron. ⁴¹ , ⁴² However, M87 and katanin are not functionally interchangeable, as evidenced by overexpression and knockdown experiments that result in different phenotypes with each of these two microtubule‐severing proteins. ³⁰ , ³² For example, experimental overexpression of M87 increases axonal branching, whereas overexpression of katanin does not. ³⁰ , ³²

Spastin preferentially severs microtubules in their stable domains, which are rich in polyglutamylated tubulin, such that inhibition of spastin results in microtubules that are abnormally long and highly polyglutamylated, and also more acetylated, given that acetylated tubulin is also enriched in the stable domains of microtubules. ⁴³ , ⁴⁴ , ⁴⁵ M1, on the contrary, is not an effective microtubule‐severing protein, but rather appears to function in the shaping of internal membranes, especially the endoplasmic reticulum (Fig. 1C). ⁴⁶ , ⁴⁷ , ⁴⁸ , ⁴⁹ , ⁵⁰ , ⁵¹ Any SPAST mutation within M87 would also be within M1, but theoretically M1 could harbor mutations not shared by M87 if the mutation occurs in those initial 86 amino acids. Over 200 different mutations to SPAST give rise to SPG4‐HSP, and these include missense, nonsense, frameshift, truncating, and splice‐site mutations, as well as deletions. ¹⁰

Observations on SPG4 ‐HSP Patients

HSPs are generally considered to be upper motor neuron diseases in which the dying‐back degeneration of corticospinal axons results in dysregulation of the lower motor neurons (also called spinal motor neurons) upon which they synapse. Why this occurs preferentially in the circuits that extend into the legs and less often to those that extend into the arms remains unknown but may relate to the greater length of the upper motor neurons relevant to the legs. Longer axons may be more vulnerable to pathological mechanisms that, for example, compromise axonal transport. In patients with more severe symptoms, it seems inescapable that there is degeneration of lower motor neurons, and therefore all patients may experience some lower motor neuron degeneration, perhaps reflective of the degree of disease progression (Fig. 2A). The paucity of genotyped HSP cadaver tissue has been an obstacle in resolving this issue. Symptoms suffered by some patients in speaking and cognition, as well as urinary and fecal performance, undoubtedly attribute to different nerve tracts (Fig. 2B), although the specifics are still unknown. Cases of SPG4‐HSP are sometimes categorized as either pure (gait defects and spasticity) or complicated (a wider spectrum of symptoms), although a growing body of evidence indicates that all cases involve symptoms that cannot be attributed solely to corticospinal degeneration.

FIG. 2.

Neural pathways involved in motor, bulbar, and autonomic control and their potential involvement in hereditary spastic paraplegias (HSPs) such as SPG4‐HSP. (A) Schematic representation of upper motor neuron (UMN) pathways from the motor cortex to spinal lower motor neurons controlling skeletal muscle. In the healthy condition, intact UMN projections descend through the brainstem to spinal motor neurons to regulate voluntary movement. Progressive UMN degeneration, characteristic of HSP, disrupts these descending signals. Early degeneration produces impaired cortical control of spinal circuits, whereas more severe degeneration leads to pronounced motor dysfunction. In patients with more severe symptoms, there is almost certainly degeneration of lower motor neurons, and therefore, all patients may experience some lower motor neuron degeneration, perhaps reflective of the degree of disease progression (A). Color code: blue, UMNs; green, lower motor neurons. (B) Neural control of bladder function, with tracts labeled on the schematic. Cortical and brainstem centers regulate urinary storage and voiding through descending pathways to the pontine micturition center and spinal autonomic circuits. Parasympathetic fibers stimulate detrusor muscle contraction, whereas somatic motor fibers control the external urethral sphincter via the pudendal nerve. Disruption of descending control in HSP can lead to bladder dysfunction, including urgency and impaired coordination between detrusor contraction and sphincter relaxation. No specific tract degeneration is shown in the schematic because any of the different tracts could be implicated in the symptoms. [Color figure can be viewed at wileyonlinelibrary.com]

The overall prevalence of SPG4‐HSP is thought to be roughly 1 in 50,000 people, with an estimated 8000 confirmed cases in the USA, although this is undoubtedly an underestimate, given the frequency of misdiagnosis. Recent imaging studies on patients are consistent with abnormalities in multiple nerve tracts, both in the brain and in spinal cord, ⁵² , ⁵³ but whether correlations exist between particular nerve tracts and specific SPAST mutations remains unclear. Gait defects in patients typically lead to wheelchair dependence. Both men and women with SPG4‐HSP (and other forms of HSP) commonly complain of urinary bladder issues. ⁶ Mild cognitive impairment, particularly in executive function, attention, and memory, is not uncommon in patients across different mutations, especially with age. ⁵⁴ When mutations occur de novo, the age of onset is typically far earlier and the symptoms far more severe than when the mutation is inherited. ⁷ , ⁵⁵ Many people with inherited mutations do not suffer symptoms until decades of symptom‐free life, whereas people with de novo mutations may already be confined to a wheelchair and suffer complex symptoms in the first few years of life. ¹⁰ In particular, the p.Arg499His mutation has appeared de novo in at least 22 documented pediatric cases of SPG4‐HSP. ⁵⁶

Recently, the most extensive clinical study to date on phenotype–genotype correlations was conducted on 206 patients with pathogenic SPAST mutations from multiple international centers. ⁵⁷ Standardized motor scales, spasticity ratings, developmental milestones, and patient‐reported outcomes were combined with plasma neurofilament assessments as a biomarker. Missense mutations were found to be more common among de novo cases, whereas inherited cases showed a breadth of mutation types but were more commonly truncated variants. Two latent disease trajectories were revealed by longitudinal analyses, one rapidly progressing (in which the de novo missense cases were enriched) and the other moderate and biphasic (in which the truncating and inherited cases were enriched). Plasma neurofilament levels were elevated for both but higher for the rapidly progressing trajectory. These results are consistent with previously reported trends and provide a basis for predicting the disease trajectory of individual patients and the potential benefits of therapies.

Disease Mechanisms

One possible disease mechanism is haploinsufficiency wherein the nerve degeneration is caused by insufficient levels of functional spastin. Another possibility is that the disease is caused by the gain‐of‐function toxicity of the mutant spastin proteins in which the mutant proteins are not only dysfunctional but acquire properties that are deleterious to cellular pathways. A third possibility is that the mutant proteins impose dominant‐negative effects that reduce the functionality of the wild‐type protein. This third possibility is especially relevant to mutant forms of M87 that are able to co‐hexamerize with wild‐type M87 and thereby reduce its activity. This would exclude truncating mutations that lack the component parts to participate in hexamer formation. Whether M1 functions as a hexamer is unknown, as is whether mutant forms of M1 can impose dominant‐negative effects on wild‐type M1. Theoretically, the disease could be caused by haploinsufficiency of M1, haploinsufficiency of M87, toxicity of mutant M1, toxicity of mutant M87, dominant‐negative effects of M87 and perhaps M1, or any combination of these possibilities. Alternatively, corticospinal neurons could be particularly vulnerable to perturbation, such that any of these mechanisms could contribute to their degeneration. This latter possibility is consistent with the growing number of other genes whose mutations cause HSP, although not easily reconcilable with the degeneration of other nerve tracts suffered by many patients.

The idea that SPG4‐HSP is caused by haploinsufficiency was originally based on the prediction that some mutations lead to nonsense‐mediated decay of either the mRNA, the protein, or both. ⁵⁸ With no mutant protein present, the reasoning was that the disease could not be caused by toxicity of the mutant protein in those cases and, therefore, was unlikely to be caused by mutant toxicity in any other case. This view was buoyed by the lack of detection of mutant spastin proteins in blood cells of patients. However, heterozygous spastin knockout mice, which model haploinsufficiency, generally do not exhibit significant nerve tract degeneration and display only mild or late‐onset motor abnormalities. Moreover, blood cells would later prove to be wholly inappropriate proxies for the cells of the afflicted nerve tracts, in part because blood cells may not express M1 at all. In fact, recent experimental work on mutations previously believed to be prone to nonsense‐mediated decay has shown the presence of both the mRNA and truncated spastin proteins in neurons. ⁵⁹

Even if nonsense‐mediated decay occurred, it could be counteracted by vanishingly low levels of the mutant protein being extremely toxic and/or accumulating due to resistance to degradation. In support of this scenario, we conducted experiments in which we introduced comparable levels of human mutant M1 or M87, bearing the same mutation, into various experimental models, including cultured vertebrate fibroblasts and neurons, squid giant axon, and Drosophila. ²⁰ , ⁶⁰ , ⁶¹ , ⁶² In all cases, and with multiple different mutations, we found mutant M1 to be toxic but its mutant M87 counterpart to be relatively harmless. In further support of our thinking, mutant M1 was more resistant to degradation than its wild‐type counterpart, whereas mutant M87 was less resistant to degradation than its wild‐type counterpart. Rapid degradation of mutant M87 would diminish its potential to act in dominant‐negative fashion.

Several years ago, we generated a novel mouse, called SPAST‐C448Y, in which we knocked in human mutant SPAST complementary DNA (cDNA), using the Rosa locus. ²¹ We chose the C448Y mutation in part because experimental studies on cells suggested that it does not have dominant‐negative properties (even though theoretically it could), at least not with regard to the microtubule‐severing properties of wild‐type M87. With both mouse SPAST genes intact, this mouse displayed adult‐onset gait defects and corticospinal dieback degeneration reminiscent of the human disease. Interestingly, the phenotype worsened in the offspring of mice produced by crossing this mouse with the SPAST knockout mouse. ²² Additional observations argue against a dominant‐negative mechanism underlying the pathology, for example, with the microtubules of the afflicted tracts showing diminished acetylation as opposed to increased acetylation, which would be predicted on the basis of lowered M87 function. ⁴⁴ The diminished microtubule acetylation shown in this animal, and more recently in human‐induced pluripotent stem cell (hiPSC)‐derived neurons with multiple different pathogenic SPAST mutations, ⁵⁹ was shown to be due to the increased activation of HDAC6. Collectively, these results support the hypothesis that toxic mutant spastin is necessary and sufficient to result in SPG4‐HSP, but that reduced spastin levels render the afflicted nerve tracts more vulnerable to the toxicity of the mutant protein. ²¹

Knockout mice are not without any phenotype as they age, but the method for generating those mice was to introduce a premature stop codon that would theoretically allow for an M1 fragment to be expressed. ⁴⁴ , ⁶³ , ⁶⁴ , ⁶⁵ Curiously, if expressed, this fragment is similar, albeit not quite as long, to an experimental peptide we previously termed M1‐STOP, which we used to demonstrate the toxic properties of a truncated M1. ²⁰ Thus, what symptoms these mice display could be due to that M1 fragment, and the same could be true of the exacerbated effect that we observed when we crossed this mouse with our genetic mouse.

In studies on squid axoplasm and cultured vertebrate fibroblasts and neurons, we documented that mutant M1 causes defects in organelle transport. ⁶² , ⁶⁶ Moreover, we found that these defects were due to aberrant phosphorylation of molecular motor proteins, which was caused by hyperactivity of casein kinase 2 (CK2). Both diminished motor protein activity and diminished microtubule acetylation could contribute to the diminished efficiency of axonal transport in the axon, which could, in turn, explain why the axon degenerates in SPG4‐HSP (Fig. 3A). Aberrations in motor protein function might also cause the accumulation of incorrectly oriented microtubules, termed microtubule polarity flaws, which in turn could cause organelle traffic jams. ⁶⁷ However, a cause‐and‐effect relationship on this has not been established; both HDAC6 and CK2 have myriad other substrates that could be affected and contribute to the axon's degeneration. Moreover, there could be entirely different reasons why mutant M1 is toxic, more important targets of its toxicity, and functional consequences more important than impaired axonal transport (Fig. 3A).

FIG. 3.

Gain‐of‐function and loss‐of‐function mechanisms potentially contributing to SPG4 hereditary spastic paraplegias (HSP). (A) Proposed pathogenic gain‐of‐function toxic cascade of mutant M1 in SPG4‐HSP. Mutant M1 drives increased casein kinase 2 (CK2) activity, which in turn causes aberrant activity of molecular motor proteins, elevates HDAC6 activity, and may contribute directly to microtubule polarity defects. Aberrant molecular motor function and microtubule polarity flaws lead to reduced axonal transport efficiency, whereas HDAC6‐mediated activity results in reduced microtubule acetylation. Notably, microtubule polarity flaws and reduced microtubule acetylation themselves may further impair axonal transport, creating a feed‐forward effect. Collectively, these molecular and cytoskeletal disruptions converge to impair axonal transport efficiency and culminate in axonal dieback degeneration. In addition, mutant M1 may directly contribute to reduced axonal transport and axonal dieback via mechanisms that remain to be elucidated. Arrows in the schematic indicate the direction of causality from mutant M1 through molecular intermediates to cellular and axonal pathology. (B) Insufficient functional M1 may result in altered endoplasmic reticulum morphology, including reduced tubulation, abnormal tubule organization, expansion of endoplasmic reticulum cisternae, and disrupted endoplasmic reticulum network continuity. (C) Insufficient functional M87 potentially leads to abnormally long and stable microtubules that are abnormally highly acetylated and polyglutamylated. [Color figure can be viewed at wileyonlinelibrary.com]

There is precedent for a domain of a protein to become toxic when the protein is not appropriately folded or otherwise regulated. In the case of tauopathies, a domain of tau called phosphatase‐activating domain (PAD) becomes toxic when mutations or other aberrations in the protein cause this normally tightly regulated domain to become constitutively active. ⁶⁸ In the case of amyotrophic lateral sclerosis (ALS), SOD1 becomes toxic when a single mutation occurs almost anywhere in the sequence. ⁶⁹ As indicated earlier, M1‐STOP peptides (280 amino acids) are toxic, ²⁰ suggesting that something in the rest of the protein is necessary for mitigating that toxicity. The hydrophobic region of M1 presumably causes it to misfold and aggregate, thereby rendering another region of M1, its toxic region, constitutively accessible.

As for the impact of haploinsufficiency, reduced spastin expression is likely to impair microtubule severing (via reduced M87) and lead to broader defects in intracellular membrane organization (via reduced M1) (Fig. 3B and C). These effects would be exacerbated if the mutant spastin proteins have dominant‐negative effects. As indicated earlier, M87 is not interchangeable with katanin, so its functions cannot be explained simply by severing any available microtubule in quantitative fashion, and this conclusion is fortified by studies showing that spastin's capacity to promote axonal growth is dependent on more than just its microtubule‐severing properties. ⁷⁰ M1's functions are even more poorly understood, with some evidence suggesting, for example, that it regulates calcium stores. ⁴⁸ Omics studies have shown that reductions in the expression of one gene can have a cascade of effects on the expression of hundreds of other genes, and some evidence suggests that this is the case with spastin. For example, one study suggests that reduced spastin levels in patient‐derived cells are accompanied by an increase in the expression of stathmin. ⁷¹ Interestingly, the precise “dosage” of spastin has been shown in Drosophila to be essential for whether amputated axons can regenerate. ⁷² Diminished regenerative capacity may reflect the same vulnerability of axons with reduced spastin to overcome the toxicity of mutant M1, but whether this vulnerability relates to M1 or M87 or both remains unknown. In an interesting study on zebrafish, different phenotypes were obtained when M1 or M87 was selectively knocked down. ⁷³

Some cases of SPG4‐HSP involve the deletion of the entire SPAST gene, which would eliminate the possibility of toxic M1, but such deletions can impact other genes as well as SPAST. ⁷⁴ , ⁷⁵ , ⁷⁶ Cases that involve mutations within the M1‐specific region that substantially alter its sequence would also not likely result in a toxic M1 protein, even a small fragment, and therefore may be the least likely to involve a gain‐of‐function etiology. ⁷⁷ What confounds understanding these cases even more is the fact that studies on large sequencing cohorts indicate that as high as 40% of HSP cases lack a known single‐gene etiology. ⁷⁸ Thus, there may well be a significant portion of SPG4‐HSP cases that are oligogenic rather than monogenic, meaning more than one gene is involved. SPG7‐HSP provides precedent for this possibility, with alterations in other genes sometimes contributing to the disease in addition to mutations in the SPG7 gene. ⁷⁹

Exon deletions can also occur, some of which remove important functional domains of spastin but are otherwise in‐frame. ⁷⁵ An in‐frame exon deletion may or may not cause a protein to become toxic, but in the case of M1, it probably does, as illustrated by an elderly human patient with a notable accumulation of the truncated M1 protein in his spinal cord. ⁸⁰ In that patient and the healthy control, even the wild‐type M1 levels were substantially higher than in cellular and animal models, suggesting either that M1 levels rise as neurons age, or that M1 is more abundant in humans than the experimental models. This might also help explain why SPG4‐HSP is a multitract disease if M1 levels are also higher in other tracts in humans.

To add even more complexity, there are cases in which a polymorphism in one SPAST gene does not cause SPG4‐HSP on its own but exacerbates the disease caused by a mutation in the other SPAST gene. ⁸¹ This may be due to the polymorphism diminishing the activity of the wild‐type protein in dominant‐negative fashion, thus making the nerve tracts more vulnerable to the impact of the pathogenic mutant protein. Rare individuals exist with biallelic variants, and they display particularly severe phenotypes. ⁵⁷

Genetic Background

The extreme variability with which symptoms become apparent in people with the same mutation suggests that genetic background is a significant contributor to SPG4‐HSP. Accordingly, a GWAS has begun to identify genetic modifiers that influence the age of symptom onset. ⁸² Genetic background probably contributes to why de novo cases (across multiple mutations, not just the ones predicted for other reasons to have more severe symptoms) tend to result in earlier and more severe phenotypes. Individuals who can better tolerate the mutation, due to genetic factors in their background, will likely have children and perpetuate the mutation, whereas individuals who cannot tolerate the mutation as well will likely never have children. These ideas are being explored in brain organoids (developed from hiPSC lines) enriched for corticospinal motor neurons, in which different SPG4‐HSP mutations were introduced against the identical genetic background. ⁵⁹ Another approach is to introduce the same mutation in cells with different genetic backgrounds to ascertain if a particular genetic background might be protective against disease severity.

Therapies for SPG4‐ HSP

We recently reported the first preclinical investigation using silence‐and‐replace gene therapy for SPG4‐HSP. ²³ We directly injected the adeno‐associated viral vector serotype 9 (AAV9) into the central nervous system (CNS), which is more effective at transducing large areas of the brain than other routes, especially when administered intracerebroventricularly (ICV). The vector was composed of a microRNA (miRNA) to knock down endogenous SPAST expression, together with cDNA of wild‐type human SPAST (modified to be unaffected by the miRNA) and was administered into neonatal SPAST‐C448Y mice via ICV injections. The treatment successfully replaced both isoforms of endogenous spastin with healthy spastin at physiological levels and prevented the onset and progression of corticospinal degeneration and gait defects (Fig. 4).

FIG. 4.

Summary of our recent gene therapy study in SPAST‐C448Y mouse model. Left panel highlights the disease model: wild‐type and SPAST‐C448Y mice are shown, expressing either normal or mutant spastin. SPAST‐C448Y mice exhibit degeneration of corticospinal tract (CST) axons and gait deficits, whereas wild‐type mice display normal CST morphology and locomotor behavior. Central panel shows the therapeutic intervention: schematic of the AAV9‐based gene therapy vector and its delivery via intracerebroventricular (ICV) injection in neonatal animals. The panel also illustrates the experimental timeline, including the schedule of behavioral and histological assessments. Right panel illustrates the conclusions: after SPAST‐AAV9‐mediated therapy, treated SPAST‐C448Y mice show intact CST axons and normal gait, indicating effective prevention of degeneration and associated deficits in motor function. [Color figure can be viewed at wileyonlinelibrary.com]

Although these findings provide encouraging proof of principle for the use of gene therapy to treat SPG4‐HSP, significant challenges remain in translating this approach to human patients. The human brain and even the adult mouse brain, being much larger than the newborn mouse brain, are challenging in terms of transduction efficiency with traditional viral vectors such as AAV9. One potential solution is an AAV capsid termed BI‐hTFR1, which binds to the human transferrin receptor (TfR1) located on the blood–brain barrier. Using mice that express TrR1, it was recently shown that this capsid produces manyfold higher transduction efficiency than AAV9. ⁸³

Expression levels are another potential issue. Overexpression of M87 would certainly have deleterious effects, and expression of M1 might as well. In our studies on the mouse model, at least at 3 and 6 months, the levels of expression from the vector never exceeded normal wild‐type levels, but whether this would be the case in humans over prolonged time spans remains unknown. New developments in gene editing, including development of base editors and prime editors, have opened the door to correcting the mutation in the patient's own DNA, which theoretically overcomes some of the potential concerns with the silence‐and‐replace approach. ⁸⁴

If the amount of M1 and M87 expressed by the one healthy gene in humans is sufficient, then the better approach might be allele‐specific antisense oligonucleotides (ASOs) that turn off expression from the mutant gene but not the healthy one. ⁸⁵ ASOs take on a widespread distribution and do not require a viral vector. ASOs are also less risky because they have a finite lifespan, after which they can be replenished if all signs are positive. On the contrary, if the pathogenic mechanism is haploinsufficiency, the gene therapy vector needs only the cDNA and not the miRNA. ASOs would not be helpful for a purely haploinsufficiency etiology but would be helpful if the mutant spastin imposed a dominant‐negative effect to compromise the activity of the wild‐type spastin.

If SPG4‐HSP is primarily based on the toxicity of M1, a limitation of the gene therapy approach is that turning off the expression of endogenous M1 may not lead to the loss of the mutant M1 that has already accumulated (and has become resistant to degradation) in the afflicted nerve tracts. Other strategies may be necessary, either together with the gene therapy or on their own, to either mitigate the toxicity of mutant M1 or, better yet, promote its degradation. One such strategy is to engineer recombinant monoclonal antibodies to the M1‐specific sequence into intrabodies that bind to the mutant M1 and promote its degradation. Intrabodies are antibodies expressed inside the cell via viral transduction. ⁸⁶ , ⁸⁷ Intrabodies to M1 could mitigate the toxicity of mutant M1 by solubilizing the M1 aggregates, by blocking the M1 toxic domains, or, if necessary, sequence can be added to target the M1–intrabody complex to the proteosome for degradation. If a potential impact on wild‐type M1 poses a problem, new monoclonal antibodies against the mutated region of mutant M1 can be generated and made into intrabodies.

As for mitigating the toxicity of mutant M1, targeting downstream effectors such CK2 or HDAC6 may be a viable strategy; however, available inhibitors are often promiscuous, and these proteins have multiple cellular targets. Targets downstream of these may be better prospects. Alternatively, existing drugs, potentially already FDA approved for other diseases, may exhibit beneficial effects, and the human organoid system offers a powerful platform to screen such compounds.

As for haploinsufficiency (and dominant‐negative disease mechanisms), recent studies suggest strategies by which the levels or activity of the available wild‐type spastin could be augmented. In one study, it was shown that 14‐3‐3 interacts with spastin via phosphorylation of spastin at Ser233, which augments the microtubule‐severing capacity of M87 and also protects spastin from degradation by inhibiting the ubiquitination pathway. ⁸⁸ Correspondingly, the 14‐3‐3 agonist fusicoccin was shown to promote spastin activity. In another study, spastin was shown to interact with UCHL1 such that its overexpression decreased the levels of spastin, whereas its knockdown increased the levels of spastin. ⁸⁹ It was also shown that the proteasome inhibitor MG‐132 inhibited the degradation of spastin by this pathway. The relevant kinase, CaMKIIβ, has been identified and could be targeted for therapeutic manipulation. ⁹⁰ In another study, murine double minute 2 (MDM2), an E3 ubiquitin ligase, was shown to interact with spastin, and that an inhibitor of MDMN2 called Nutlin‐3a upregulates spastin functions. ⁹¹ Finally, increasing spastin activity may be possible by reducing the levels or activity of fidgetin, another microtubule‐severing protein that may co‐hexamerize with M87 to tamp down its activity. ⁹² , ⁹³ All of these studies offer druggable pathways to increase spastin activity, although it is unclear whether the results apply to M87 and M1 or only to M87.

Reversal of Symptoms

The various therapies discussed thus far all share the limitation that they probably cannot reverse the nerve tract degeneration that has already occurred or the symptoms associated with that degeneration. The younger the patient the more likely the nerve tracts might resume normal development, but for any patient, synaptic connections that have been lost are unlikely to be restored as a result of the degeneration being halted. Thus, for the vast majority of sufferers, including many whose nerve tracts have been degenerating for decades, therapies that only slow the progression of the disease are wholly unsatisfactory.

Noteworthy efforts in the spinal cord injury (SCI) field have focused on therapies that cause sprouting of healthy axons such that the new sprouts can make appropriate synaptic connections to enable functional recovery. ⁹⁴ In some cases, axons that were degenerating may be able to generate sprouts (ie, collateral branches) that form appropriate synaptic connections. Over the past few decades, exercise has proven to be the most effective means for achieving these benefits. The formation of new axonal branches and circuits is mostly attributable to the enhancement of secretion of neurotrophic factors, but there are also neuroprotective and restorative effects on damaged nervous tissue when exercise (such as treadmill) is utilized following SCI and other CNS injuries and diseases. ⁹⁵ These benefits, both in animal models and in human patients, have been shown to promote cognitive function recovery, reverse neuroinflammation, reduce neuronal apoptosis, and enhance markers of neuroplasticity. Especially relevant to SPG4‐HSP, exercise has also been shown to enhance the autophagic degradation of misfolded proteins in other diseases. ⁹⁶

Promise has also been shown in the use of spinal cord stimulation (SCS) to reduce spasticity and restore function across a variety of neurological disorders. Noninvasive transcutaneous SCS (tSCS) has been shown to improve both motor and autonomic function in individuals with SCI and cerebral palsy. ⁹⁷ , ⁹⁸ Encouragingly, case reports have shown that invasive epidural SCS (eSCS) can also improve motor and autonomic functions in individuals with HSP, suggesting that spinal circuits in patients with HSP remain responsive to neuromodulation. ⁹⁹ , ¹⁰⁰ , ¹⁰¹ , ¹⁰²

We posit that these strategies will enable the formation of new circuits and connections sufficient to support at least partial functional recovery in SPG4‐HSP patients once the therapies discussed earlier have halted disease progression. We prefer this pro‐regenerative approach to selective dorsal rhizotomy (SDR), while acknowledging that SDR has provided symptom improvement in some patients. ¹⁰³

The Need for More Knowledge and Tools

Given the risks associated with any therapy, treating asymptomatic individuals may be premature, particularly when symptom onset may be delayed for decades. Thus, a reliable blood‐based biomarker for disease progression is critical for introducing potential therapies at the best time. In addition, the most useful biomarker would be the one whose circulating levels correlate with disease progression and demonstrate responsiveness to treatment. At present, neurofilament protein is being used as a biomarker for HSP, ⁸² , ¹⁰⁴ , ¹⁰⁵ but a protein (or protein variant such as a phospho‐variant) more reflective of the specific tracts that degenerate is sorely needed.

The current mouse models are useful, and yet we can do better in making a model that reflects more aspects of the human disease. Ideally, such a mouse would replace both mouse SPAST genes with human SPAST genes, one of which would bear a pathogenic mutation. A mouse with a mutation introduced into an endogenous SPAST gene has an extremely mild phenotype (unless the mutation is homozygous), suggesting that the human gene may be critical for modeling the disease. ¹⁰⁶ Although the technology for gene editing is not as advanced in animals other than mice, there may be value in considering animal models (such as rat) that are more amenable to certain kinds of studies, such as neurostimulation. Important to keep in mind is the fact that mice and rats are anatomically different from humans, for example, with regard to the tracts relevant to gait, and therefore interpreting results with these model animals requires due caution. As discussed earlier, there is power in hiPSC‐derived cells that can be differentiated into neurons and multicellular organoids, ³³ , ⁵⁹ , ¹⁰⁷ , ¹⁰⁸ , ¹⁰⁹ , ¹¹⁰ , ¹¹¹ , ¹¹² , ¹¹³ , ¹¹⁴ , ¹¹⁵ but of course they cannot be used for analyses of behavioral symptoms such as gait.

Concluding Remarks

In conclusion, we are pleased to share this status update on current understanding of SPG4‐HSP, with emphasis on the outlook for development of therapies (see Table 1). Although there are multiple obstacles to overcome, there is also far better clarity on this disease than there was just a few years ago. We are hopeful that more scientists and clinicians will join our cause in conquering this disease.

TABLE 1.

Summary of potential therapies for SPG4 hereditary spastic paraplegia (HSP), highlighting the strengths, as well as potential shortcomings and obstacles, of each one

Therapy	How does it work?	Strengths	Potential obstacles
Silence‐and‐replace gene therapy	Turns off expression of endogenous SPAST genes, both mutant and wild type, and replaces them with expression from wild‐type SPAST cDNA	Treats both loss‐of‐function and gain‐of‐function components of the diseases	Optimal virus for transduction of human brain is still under development
			Unclear whether enough neurons in human brain would be transduced
			Too much M87 expression would likely kill the transduced neurons
			Too much M1 expression would likely exacerbate the disease
			Mutant M1 that has accumulated is likely resistant to degradation, even when expression of more is halted
			Probably limited to slowing or halting disease progression rather than restoring lost function
Allele‐specific antisense oligonucleotides	Turns off expression only of the mutant SPAST gene	Does not require viral transduction	Treats gain‐of‐function and dominant‐negative components of the disease but not haploinsufficiency
			Requires one functional SPAST gene to be enough
		Widespread distribution
			Usually designed from individual patient's intronic sequence, so they are patient‐specific and expensive
		Degrades over time (requiring regular boosters) so problems that might arise are likely to be reversible
			Probably limited to slowing or halting disease progression rather than restoring lost function
M1‐specific intrabodies	Deaggregates, detoxifies, and/or degrades mutant M1	Directly switches off the culprit of the gain‐of‐function etiology of the disease, regardless of culprit's increased resistance to degradation	Optimal virus for transduction of human brain is still under development
			Unclear whether enough neurons in human brain would be transduced
			Only treats gain‐of‐function component of the disease
			Requires one functional SPAST gene to be enough
			Could inhibit or degrade wild‐type M1 as well as mutant M1, leading to potential M1 deficiency
			Probably limited to slowing or halting disease progression rather than restoring lost function
Drugs	Some drugs could increase the activity of wild‐type spastin, and other drugs could inhibit downstream pathogenic pathways of mutant M1	Easy for patient to take with outcomes rapidly known and any potential ill effects likely reversible upon cessation of use	Probably limited to slowing or halting disease progression rather than restoring lost function
Exercise and other physical therapy modalities, neurostimulation	When disease mechanisms are slowed or halted by one or more of the other therapies, exercise and other physical therapy modalities, as well as neurostimulation, can stimulate the formation of new axonal branches and circuits that can restore lost function	The only therapy that likely restores lost function rather than just slowing or halting disease progression if coupled to one or more of the other therapies	On their own, have limited benefit beyond making symptoms more tolerable
			Likely need to be coupled with one of the other therapies to elicit genuine restoration of lost function
		Has a variety of other benefits, such as maintaining muscle tone
			Likely do not restore the same circuits that are lost but rather establish new alternate ones, thus not perfectly restoring the functions that have been lost

Abbreviation: cDNA, complementary DNA.

Author Roles

The authors (Piermarini and Baas) conceived and wrote the article and prepared the display items collaboratively.

Financial Disclosures of All Authors (for the Preceding 12 Months)

None.

Financial Disclosures and Conflicts of Interest

Author disclosures are available in the Supporting Information.

Supporting information

Supporting Information.

Data Availability Statement

Data sharing not applicable to this article as no datasets were generated or analysed during the current study.