Biomarkers for predicting disease course in Sanfilippo syndrome: An urgent unmet need in childhood‐onset dementia

Sanfilippo syndrome (MPS III) is an autosomal recessive inherited disorder causing dementia in children, following an essentially normal early developmental period. First symptoms typically include delayed language development, hyperactivity and/or insomnia from 2 years of age, followed by unremitting and overt loss of previously acquired skills. There are no approved treatments, and the median age of death is 18 years. Treatments under clinical trial demonstrate therapeutic benefit when applied pre‐symptomatically in children diagnosed early through known familial inheritance risk. Newborn screening for Sanfilippo syndrome would enable pre‐symptomatic diagnosis and optimal therapeutic benefit, however, many fold more patients with Sanfilippo syndrome are expected to be identified in the population than present with childhood dementia. Therefore, the capacity to stratify which Sanfilippo infants will need treatment in toddlerhood is necessary. While diagnostic methods have been developed, and continue to be refined, currently there are no tools or laboratory‐based biomarkers available to provide pre‐symptomatic prognosis. There is also a lack of progression and neurocognitive response‐to‐treatment biomarkers; disease stage and rate of progression are currently determined by age at symptom onset, loss of cerebral grey matter volume by magnetic resonance imaging and developmental quotient score for age. Robust blood‐based biomarkers are an urgent unmet need. In this review, we discuss the development of biomarker assays for Sanfilippo based on the neuropathological pathways known to change leading into symptom onset and progression, and their performance as biomarkers in other neurodegenerative diseases. We propose that neural‐derived exosomes extracted from blood may provide an ideal liquid biopsy to detect reductions in synaptic protein availability, and mitochondrial function. Furthermore, given the prominent role of neuroinflammation in symptom expression, glial fibrillary acidic protein detection in plasma/serum, alongside measurement of active brain atrophy by neurofilament light chain, warrant increased investigation for prognostic, progression and neurocognitive response‐to‐treatment biomarker potential in Sanfilippo syndrome and potentially other childhood dementias.


| SANFILIPP O SYNDROME , CLINIC AL PRE S ENTATI ON AND CURRENT TRE ATMENTS
Sanfilippo syndrome, otherwise known as mucopolysaccharidosis type III (MPS III), is an autosomal recessively-inherited disorder characterised by the inability of the lysosome to degrade the mucopolysaccharide, heparan sulfate. Heparan sulfate degradation occurs in a stepwise manner. Mutations in any one of the four genes encoding enzymes in this pathway lead to complete loss of protein expression or a significant reduction in the specific enzyme's activity and Sanfilippo syndrome subtypes A-D (Meikle et al., 1999) and E (Kowalewski et al., 2012). As partially degraded heparan sulfate accumulates in lysosomes of most cells of the body, massive storage vacuoles form. The central nervous system (CNS) is the most severely affected organ (Neufeld & Muenzer, 2001;Witting et al., 1975).
Despite heparan sulfate accumulation commencing in utero, children with MPS III have an essentially normal early developmental period until approximately 2 years of age, from which point cumulative and progressive loss of previously acquired skills causing dementia unfolds. Sanfilippo patients are typically diagnosed from 3 to 6 years of age (MPS III types A, B or D) or 10 to 12 years of age for MPS III type C (Héron et al., 2011;Martins et al., 2019;Meikle et al., 1999); to date, no MPS III type E patients have been diagnosed in childhood, rather these patients have so far presented with adult-onset sensorineural hearing loss and retinitis pigmentosa (Usher syndrome type 4; Khateb et al., 2018;Velde et al., 2022), although selfreported childhood-onset sensorineural hearing loss is described in this patient group (Velde et al., 2022). Initial misdiagnosis of Sanfilippo syndrome is not uncommon given that the first cognitive symptoms typically include delayed language acquisition, followed by the onset of abnormal behaviour (hyperactivity, irritability, aggressiveness, lack of fear, poor attention and autistic-like symptomatology in some individuals; Héron et al., 2011;Martins et al., 2019;. As the disease progresses, insomnia, loss of acquired speech and severe neurocognitive regression ensue. Children lose independent walking, relational interactions and progress to a vegetative state followed by death often in the second or third decades of life (Héron et al., 2011;Martins et al., 2019;Ruijter et al., 2008;Valstar et al., 2008;van de Kamp et al., 1981).
Although there is no approved treatment available for Sanfilippo syndrome, many clinical trials aimed at restoring functional enzyme activity to the brain, either by direct delivery of recombinant en-  (Flanigan et al., 2021), and study follow-up continues. In addition, a gene therapy study in MPS IIIB patients also explicitly suggests children less than 24 months of age should be treated for optimal therapeutic outcomes (Tardieu et al., 2017). Similar results have been found for other treatments at clinical trial for Sanfilippo syndrome (Kohn et al., 2020;Tardieu et al., 2014;Wijburg et al., 2019). Moreover, animal studies using the same treatments support these findings, with improved behavioural outcomes following pre-symptomatic compared to symptomatic treatment protocols (Cressant et al., 2004;Fraldi et al., 2007;Fu et al., 2016;Hemsley et al., 2009;Savas et al., 2004).
Thus, there is a need to identify patients pre-symptomatically to offer well-timed treatment, for optimal therapeutic outcomes.
Newborn screening (NBS) for Sanfilippo syndrome is not currently a routine practice in any country of the world. Rather, beyond known familial inheritance risk, diagnoses are most often made at the symptomatic stage. Given the limited window for optimal treatment initiation, inclusion of Sanfilippo syndrome into NBS programmes will be necessary.

| NBS FOR SANFILIPP O SYNDROME
Newborn screening, performed on several drops of blood from a 2-to 3-day-old infant, is used to identify infants at risk of developing fatal or disabling conditions prior to symptom onset, optimising treatment/therapy efficacy and limiting adverse outcomes for the infant. NBS for Sanfilippo syndrome would enable asymptomatic access to a clinical trial-or then proven-treatment, optimising treatment outcome and reducing morbidity and mortality. Additionally, families could obtain genetic counselling. This is particularly relevant as many families are faced with clinical presentation of their eldest affected child, after subsequent birth(s) (Héron et al., 2011), and many families have multiple severely affected children. response-to-treatment biomarker potential in Sanfilippo syndrome and potentially other childhood dementias.

K E Y W O R D S
biomarker, childhood dementia, lysosomal storage disease, MPS III, prognostic, Sanfilippo Unfortunately, broad NBS for neurodegenerative lysosomal storage disorders (LSDs) does not occur in any country around the world (Taiwan and states of the USA are the fastest moving in the field, Jalal et al., 2022). Furthermore, NBS is anticipated to identify significantly more patients with Sanfilippo syndrome than those presenting clinically with childhood-onset dementia (for whom early treatment is crucial), as some MPS III patients have been found to present in adulthood with retinal degeneration and/or cardiomyopathy, with or without neurodegeneration (Berger-Plantinga et al., 2004;Nijmeijer et al., 2019;Van Hove et al., 2003). Furthermore, NBS will reveal novel mutations, mutations causing partial enzyme deficiency and common mutations alongside currently unknown epigenetic factors; leaving unclear the clinical significance of an NBS diagnosis without further follow-up assessments over time. This is a common problem in other LSDs. Prior to NBS, LSD prevalence in the population was determined by the number of clinical diagnoses compared to population births. By this method, the prevalence of Fabry, Gaucher and Krabbe diseases, for instance, has been described, ranging from 0.71 to 1.75 per 100 000 births (Meikle et al., 1999). With the implementation of NBS, the actual prevalence of these disorders detected in the population has increased dramatically: 12-to 37-fold for Fabry disease (Elliott et al., 2016;Spada et al., 2006;Wasserstein et al., 2019), 4-to 13-fold for Gaucher disease (Elliott et al., 2016;Wasserstein et al., 2019) and 35-fold for Krabbe disease (Elliott et al., 2016) depending on the population. This suggests a high percentage of infants identified with the biochemical and genetic traits of Sanfilippo syndrome through NBS are likely to present later in life, potentially with non-neurological symptoms (Nijmeijer et al., 2019), if they present at all. Therefore, tools capable of providing prognosis following NBS are essential, to enable timely stratification of patients who will develop childhood dementia and require what is likely to be exorbitantly expensive and sometimes irreversible treatment before 2 years of age.
Furthermore, gauging where in the pre-onset trajectory Sanfilippo patients are prior to treatment is also essential, to better interpret patient response to treatment and the maximal therapeutic benefit attainable for a given treatment and treatment-protocol.
Currently, this can only be achieved with neurocognitive measures (DQ scores) using time-consuming behavioural testing performed by experts, and specialised magnetic resonance imaging (MRI) to measure grey matter volume. These measures are also the only neurocognitive response-to-treatment measures available and require extensive long-term follow-ups.
Although diagnostic biological markers (biomarkers) for Sanfilippo syndrome are well established, no broadly effective prognostic, progression or neurocognitive response-to-treatment biomarkers are available to date.

| B I OMARK ER T YPE S AND CONTE X T OF US E DEFINED
All biomarkers must perform based on their intended context of use.
Diagnostic biomarkers identify disease risk in an individual or confirm disease type following symptom onset. Prognostic biomarkers are essential to identify how a diagnosed individual will subsequently progress the disease pathway. Disease progression biomarkers are required to measure the state/phase of disease entered at followup assessments, while pharmacodynamic (response-to-treatment) biomarkers are required to measure how effective a treatment was for an individual at their treated disease stage (FDA-NIH Biomarker Working Group, 2016).
The outcomes of any clinical trial hinge in a large part on the capacity to stratify patients into appropriate clinical trial groups (using progression and prognostic biomarkers) and the subsequent ability to interpret treatment effects at a given treated disease 'stage' or 'risk of disease onset'. Neurocognitive outcome is essential for gene therapy/enzyme replacement therapy in Sanfilippo patients.
Without prognostic, progression and neurocognitive response-totreatment biomarkers to stratify patients into groups and measure outcomes, treatments may not gain regulatory approval, despite clear but inexplicable neurocognitive benefits being measured in sub-populations of patients in a treatment cohort.

| Heparan sulfate
Diagnostic biomarkers are already available for Sanfilippo syndrome including elevated urinary and blood spot, heparan sulfate and severely reduced enzyme activity because of mutation(s) in a Sanfilippo-causing gene (diagnostically applicable to newborn blood spot screening, Yi et al., 2018). Despite delayed symptom onset, heparan sulfate accumulation commences in utero in MPS IIIA in humans (Martin & Ceuterick, 1983). Using the most common (D31N) mouse model of MPS IIIA, heparan sulfate accumulation at birth was confirmed, increasing in amount until approximately 13 weeks of age (Crawley et al., 2006). Heparan sulfate levels have been assessed in Sanfilippo patients for prognostic or disease-tracking value, finding no benefit (Nijmeijer et al., 2019;. At clinical trial, heparan sulfate remains the primary laboratory-based biomarker for measuring therapeutic efficacy (Hocquemiller, 2021). While it does demonstrate the efficacy of the treatment to reduce the primary storage burden of disease, it has no efficacy as a neurocognitive response-to-treatment biomarker. For instance, while intra-CSF recombinant human enzyme replacement therapy in MPS IIIA patients provided no cognitive benefit (Clini calTr ials.gov NCT02060526), CSF heparan sulfate levels near normalised .
In contrast, AAVrh10-based intracerebral gene therapy in MPS IIIB patients (EudraCT: 2012-000856-33) demonstrated some cognitive benefit, however, high CSF heparan sulfate levels persisted (Tardieu et al., 2017). Furthermore, AAV9-based gene therapy in MPS IIIA patients led to reductions in CSF heparan sulfate equally across all patients who received the highest vector genome treatment. However, patients in this group who were treated pre-symptomatically (less than 30 months of age) were the only participants to maintain a normal/healthy developmental trajectory; the rest of the patients in that treatment group (despite having reduced CSF heparan sulfate) did not gain cognitive benefit when treated symptomatically (Kakkis & Flanigan, 2022). Overall, these studies show that CSF heparan sulfate levels cannot be used to stratify cognitive response to treatment.

| Residual enzyme activity
Residual enzyme activity has been assessed for prognostic value. It has been assumed that the Sanfilippo phenotype is the result of a graded level of residual mutant enzyme activity and the corresponding elevation in primary substrate accumulation (Meyer et al., 2008), where residual activity of 5%-10% of normal was expected to have a more attenuated (or no) phenotype . Higher residual enzyme activity is, however, not always associated with a milder phenotype or slower evolution of disease (Héron et al., 2011;Tardieu et al., 2014). Furthermore, residual enzyme activity of 0%-3.3% of the lower normal range does not guarantee a dementia phenotype in childhood or adulthood (Nijmeijer et al., 2019). Given that residual enzyme activity measures performed in vitro typically employ artificial (fluorogenic) substrates rather than the natural oligomeric substrate, they may be considered predictive rather than an absolute measure of a patient's residual enzyme activity. Overall, residual enzyme activity in Sanfilippo syndrome is not highly predictive of disease state, rate of progression or clinical course likely to be taken.

| Mutation type
Mutations have been assessed for prognostic purposes but are for the most part ineffectual in predicting rate of disease progression in Sanfilippo syndrome. Overall, hundreds of mutations in the Sanfilippo genes have been documented (Canals et al., 2011;Feldhammer et al., 2009;Jansen et al., 2007;Lee-Chen et al., 2002;Tanaka et al., 2002;Tessitore et al., 2000;Yogalingam & Hopwood, 2001) and novel mutations continue to be identified (Martins et al., 2019;Nijmeijer et al., 2019;. Instances where genotype-phenotype correlations have appeared promising include the identification of the p.S298P homozygous mutation in MPS IIIA producing a late-/adult-onset or attenuated phenotype. This mutation when presented as a compound heterozygote with a mutation frequently associated with an early-onset phenotype (such as p.R245H, p.Q380R, c.1080delC or p.S66W) results in the slowprogressing childhood-onset form of the disease (Meyer et al., 2008;Valstar et al., 2010). However, variations are often observed in the clinical courses of affected siblings who carry the same mutations (Héron et al., 2011;Martins et al., 2019;Nijmeijer et al., 2019;Valstar et al., 2008Valstar et al., , 2010van de Kamp et al., 1981), suggesting epigenetic, potentially environmental factors or unidentified polymorphisms that may influence disease expression (Perkins et al., 1999). For the most part, mutation type holds little prognostic value, particularly in the case of novel mutation identification.

| Symptomology
Symptomology has also been assessed for prognostic value, where the patient's age at diagnosis or precocity of symptom onset predicts the rate of decline. Those patients presenting before the age of 5 years tend to display more rapid disease progression, compared to those presenting later in childhood (Héron et al., 2011). Knowledge gained from natural history studies demonstrates that onset of behavioural symptoms closely associates with atrophy of the amygdala and brain generally (Potegal et al., 2013;Shapiro, King, et al., 2016;. To date, precocity of symptom onset and loss of grey matter volume (measured by MRI) and DQ are the best prognostic tools available. With grey matter volume reducing with age, and DQ reducing with grey matter volume loss (Heon-Roberts et al., 2020), these measures also effectively demonstrate response to treatment in clinical trial (Kakkis & Flanigan, 2022;Laufer, 2022;Mathers, 2022).
Although MRI shows clear prognostic outcomes in symptomatic patients, MRI has limited use in the very young because of the lack of contrast between cerebral grey and white matter (caused by early changes in grey matter water content and developing myelin), as discussed by Tardieu et al. (2017). Furthermore, the required equipment is not widely accessible throughout the world. The use of DQ as a pre-symptomatic prognostic tool unfortunately is also unviable in the very young, as it is impossible in newborns and very hard to implement in toddlers, requiring lengthy testing by highly trained specialists for robust outcomes. Consequently, there is an urgent need to develop prognostic and progression biomarkers that can be implemented in the very young. Ideally, these biomarkers need to be measurable in peripheral fluids (such as blood, saliva and urine), using testing methods accessible in clinical laboratories worldwide.
The successful discovery of biomarkers to determine risk-to-onset or progression of neurodegenerative disease, requires knowledge of the neuropathological pathways leading to Sanfilippo dementia.

| MECHANIS MS OF D IS E A S E: A SOURCE OF PROG NOS TI C AND RE S P ON S E-TO -TRE ATMENT B I OMARK ER S?
As discussed above, heparan sulfate itself provides diagnostic benefits but does not define disease trajectory (similarly to residual enzyme activity and novel mutations in prognosis); additionally investigational treatments can resolve heparan sulfate without reversing already instigated symptom onset. Based on this, neuropathological pathways triggered by heparan sulfate accumulation represent the first potential pathways/proteins with prognostic or progression biomarker value. The precise mechanism by which primary storage of heparan sulfate causes neurological impediment and decline has not been elucidated. The pathways implicated in CNS disease onset and progression have been extensively investigated, since the first description of Sanfilippo syndrome in 1963 (Sanfilippo et al., 1963). Briefly, as summarised in Figure 1, the accumulation of heparan sulfate starts a cascade of cellular mis-regulations, originating locally by heparan sulfate itself altering the activity of other lysosomal degradative enzymes. The secondary accumulation of gangliosides is an example of this (McGlynn et al., 2004). Lysosomal dysfunction progresses further intracellularly to inhibit axonal trafficking and normal autophagic pathways, allowing polyubiquitinated proteins (such F I G U R E 1 Cellular pathways of health or Sanfilippo syndrome disease development. In healthy brain, cells degrade heparan sulfate in the lysosome and homeostatic cellular functions (autophagy, neuroaxonal trafficking, synaptic vesicle docking, neurotransmitter release, miniature post-synaptic current and action potential conduction) are maintained, promoting robust neural network formation with the support of healthy astrocytes and microglia. In Sanfilippo brain, partially degraded heparan sulfate accumulates progressively in lysosomes, engorging the cell. Accumulated dysfunctional lysosomes fuse poorly with autophagosomes, inhibiting autophagy. Consequently, protein and damaged mitochondria accumulate. Neuroaxonal trafficking is inhibited, and axonal dystrophy is observed-likely contributing to the reduction in synaptic proteins observed with disease. Palmitoylation is also reduced, destabilising synaptic protein cysteine string proteinα, by targeting it to the proteasome for degradation. Consequently, overall SNARE complex formation, vesicle docking and neurotransmitter exocytosis into the synaptic cleft are reduced. At the post-synapse miniature excitatory, post-synaptic currents are reduced in frequency and amplitude making local synapses vulnerable to tagging for pruning. Simultaneously, the innate immune system is primed by heparan sulfate at toll-like receptor-4 and activated by secondary storage products at the NLRP3 inflammasome; pro-inflammatory cytokines are released and neuroinflammation is initiated. Gliosis ensues with activated astrocytes and microglia exacerbating the neuroinflammatory state and phagocytosing inactive synapses-causing the reduction in spine density observed with disease. Neuroinflammation also fosters an environment suitable for pyroptosis and the reduction in grey matter volume observed with disease. Overall, these pathogenic pathways compound, reducing brain plasticity and leading to the cognitive decline and symptom progression characteristic of Sanfilippo dementia. ATP, adenosine triphosphate; mEPSC, miniature excitatory post-synaptic currents; NLRP3, NLR (nucleotide-binding domain leucine-rich repeat-containing) family Pyrin domain-containing 3; SNARE, soluble N-ethylmaleimide-sensitive factor attachment protein receptor. Figure  created with BioRe nder.com. as tau, α-synuclein and ubiquitin; Beard et al., 2017;Heon-Roberts et al., 2020;Sambri et al., 2017) and damaged mitochondria to accumulate (Settembre et al., 2008), likely releasing reactive oxygen species and disrupting the normal energy balance in cells. In addition, reduced palmitoylation has been found to promote increased proteasomal degradation of specific proteins-reducing their time for function. As CNS pathology progresses, changes in synaptic function and structure appear, likely aided by neuroinflammation (and excessive synaptic pruning) with several studies showing (i) a reduction in synaptic densities (Pará et al., 2021;Sambri et al., 2017;Takashima et al., 1985), (ii) synaptic activity (Dwyer et al., 2017;Pará et al., 2021;Sambri et al., 2017) and (iii) an up-regulation in neuroinflammation (Hamano et al., 2008;Hassiotis et al., 2014;Jones et al., 1997;Witting et al., 1975, as reviewed by Mandolfo et al., 2022.
The following section discusses potential novel biomarkers for Sanfilippo syndrome based on known neuropathological pathways; alongside the respective proteins that have been measured in the blood of patients with other neurodegenerative diseases, where they show prognostic or progression biomarker potential, as collated in Table 1.

| Altered heparan sulfate sulfation patterns
Accumulated mucopolysaccharides have been found to stably bind lysosomal hydrolases, producing isoenzymes with increased (βglucuronidase, N-acetylβ-hexosaminidase and α-fucosidase) and decreased (α-and β-galactosidase and arylsulfatase A) activity (Kint et al., 1973). When neuraminidase is inhibited by mucopolysaccharide accumulation, subsequent accumulation of ganglioside GM3 is observed, potentially explaining its secondary accumulation in Sanfilippo, as discussed by McGlynn et al. (2004). Secondary storage of GM2 ganglioside appears to be the result of up-regulated synthesis at the Golgi (McGlynn et al., 2004). Evaluation of human post-mortem MPS IIIA, C and D patient cerebral cortex not only confirmed both GM2 and GM3 ganglioside accumulation alongside the expected elevation in heparan sulfate but also unexpectedly identified a simultaneous reduction in keratan sulfate. How heparan sulfate accumulation promotes this effect remains to be elucidated (Viana et al., 2020). Given that all mucopolysaccharidoses store mucopolysaccharides but only MPS subtypes I, II, III and VII store heparan sulfate and develop severe neurological symptoms (Bigger et al., 2018), it is reasonable to suggest that as heparan sulfate storage fills and exceeds lysosomal housing, lysosomal and extralysosomal enzymes' activities and protein functions are significantly altered by heparan sulfate itself, triggering neurodegenerative outcomes that once evolved to the symptomatic stage appear to not be easily reversed by treatments purely addressing heparan sulfate storage.
Heparan sulfate functions through its sulfation pattern. An increase in 2-O-sulfated disaccharides, at the expense of non-6-Osulfated and unmodified disaccharides in MPS IIIA and IIIB mouse brain homogenate has been observed (Wilkinson et al., 2012), likely increasing heparan sulfate's biological activity at sites of accumulation. For example, 2-O-sulfated heparan sulfate has been found to promote toll-like receptor 4 (TLR4) priming of the innate immunity, while secondarily stored products have been found to activate NLRP3-inflammasome, together promoting neuroinflammation (Parker et al., 2020, reviewed by Mandolfo et al., 2022. Although total heparan sulfate has no prognostic/progression/or neurocognitive response-to-treatment benefit, evaluation of 2-O-sulfated heparan sulfate fragment accumulation over the pre-symptomatic to symptomatic period may warrant consideration. However, measurement would need to occur in CSF in order to measure brainspecific changes.

| Autophagy and proteasomal defects
Autophagy is important for constitutive protein turnover and mitochondrial quality control. In MPS III, the intense accumulation of lysosomes packed with heparan sulfate ultimately results in inhibition of autophagosome-lysosome fusion, reducing autophagic function (Settembre et al., 2008). In a recent post-mortem study, hallmarks of a block in autophagy were detected in all Sanfilippo subtypes assessed (MPS IIIA, C and D; Viana et al., 2020). Elevated lysosomal pH has been described to inhibit autophagosome-lysosomal fusion in CHO cells previously and may play a role in MPS III (Kawai et al., 2007).
As a consequence of impaired axonal transport and/or reduced autophagic activity, proteinaceous inclusions containing ubiquitin, tau or α-synuclein are observed repeatedly in human post-mortem tissue taken from Sanfilippo patients and mouse models in the presence of an apparently normally functioning proteasomal system, as reviewed in Heon-Roberts et al. (2020). Reduced pre-synaptic palmitoylation has been found, however, to destabilise and target pre-synaptic protein cysteine string proteinα to the proteasome for degradation, reducing its functional half-life in MPS III A mice. This pre-synaptic reduction in cysteine string proteinα, together with a reduction in soluble α-synuclein at the synapse, inhibits normal soluble NSF attachment receptor (SNARE) complex formation and synaptic functioning (Sambri et al., 2017). Treatment restoring cysteine string proteinα function rescued synaptic vesicle number and ameliorated synaptic function, prolonging survival in MPS IIIA mice (Sambri et al., 2017).   Welford et al. (2022) Gliosis detected in human MPS III brain postmortem (Jones et al., 1997;Kurihara et al., 1996;Tamagawa et al., 1985). Astrocytosis and microgliosis well established by 4+ weeks of age in MPS III mice (Hassiotis et al., 2014). Age at establishment reflecting sub-type severity (Wilkinson et al., 2012) (Continues)

Biomarker protein
Neurodegenerative disease Sample

Plasma
No robust neurocognitive biomarker potential. Increased levels of pro-inflammatory cytokines detected but levels are highly influenced by general infection states Parker et al. (2020) In MPS III mouse brain increased levels of proinflammatory cytokines; inhibition of IL-1-driven inflammation prevented neurological decline (Parker et al., 2020;Wilkinson et al., 2012) Synaptic function Neurotransmitters Alzheimer's disease and frontotemporal dementia Blood BDE extracts Prognostic potential: synaptophysin, synaptopodin, synaptotagmin-2 and neurogranin (and synapsin in Alzheimer's disease only) reduced in healthy/mild cognitive impairment state, 1-10 years prior to diagnosis. Potential progression and neurocognitive response to treatment as synaptophysin, synaptotagmin and synaptopodin levels also demonstrated a relationship with cognitive performance Goetzl et al. (2016) Reductions in dendritic spines near mega-neurites and focal dendritic enlargements at human MPS III brain autopsy (Ferrer et al., 1988;Takashima et al., 1985). In MPS III mouse models, decreased synaptic vesicle trafficking to the synapse and reduction in functional synaptic proteins: synaptophysin, vesicle-associated membrane protein (VAMP-2), syntaxin-binding protein 1, synapsin-1, calcium/calmodulin-dependent protein kinase II and neuroligin-2. Reductions in: brain-derived neurotrophic factor, synaptic activity and synaptic density (Dwyer et al., 2017;Pará et al., 2021;Sambri et al., 2017;Villani et al., 2007;Vitry et al., 2009;Wilkinson et al., 2012) Cerebral

TA B L E 1 (Continued)
future cognitive decline and brain atrophy, as discussed by Blennow and Zetterberg (2018). Based on the results of many studies, single molecule array (SIMOA) blood-based phospho-Tau 181 analysis has received FDA breakthrough device status for AD diagnosis (Quanterix, 2022a, January 12a). Although clinically exciting for neurodegenerative diseases, when applied to Sanfilippo patients CSF, modest tau elevations reducing with age were detected , a behaviour likely limiting its prognostic potential in Sanfilippo patients. The CSF finding is not surprising given that neurofibrillary tangles and plaques are not a feature of Sanfilippo disease neuropathology.
Alpha-synuclein has been assessed for its biomarker potential in neurodegeneration, given it is the primary component of Lewy bodies characterising Parkinson's disease (PD). In PD patient saliva, α-synuclein levels have been reported to reduce mirroring CSF levels and correlating with PD disease stage, as reviewed by Farah et al. (2018). In blood, however, no correlation of α-synuclein was found with disease state or stage, as reviewed by Lewczuk et al. (2018). Whether this is the result of poor-quality immunoassays previously available for research purposes remains to be determined. Assay inhibition by serum/plasma matrix is oftentimes poorly assessed in tests validated and sold for research purposes. Careful use of controls is essential to ensure that the output results reflect a sensitive and specific assay. In Sanfilippo, α-synuclein has been found accumulated in axonal spheroids (Beard et al., 2017;Heon-Roberts et al., 2020;Sambri et al., 2017) and whether α-synuclein has any biomarker potential in Sanfilippo remains to be determined.

| Mitochondrial dysfunction
Corresponding to the reduction in autophagic activity in Sanfilippo, structurally abnormal mitochondria, along with impaired mitochondrial activity, have been measured in MPS IIIC mice (Martins et al., 2015). Mitochondrial budding has also been noted in Sanfilippo human post-mortem brain tissue (Haust, 1968). Furthermore, in MPS IIIC mouse brain synaptosomes, reduced levels of mitochondrial proteins were detected, compared with control (Pará et al., 2021).
Whether insufficient autophagic activity is solely responsible for the mitochondrial deficits in MPS III remains to be determined. However, a similar investigation using a PD patient cell line caused by haploinsufficiency of the lysosomal enzyme β-glucocerebrosidase (an enzyme that when completely insufficient results in the childhood neurodegenerative condition: Gaucher's disease) found prolonged lysosomal-mitochondrial contact tethering and defects in mitochondrial distribution and function (Kim et al., 2021). Lysosomalmitochondrial contact tethering is important in calcium, iron and cholesterol transfer between the two organelles (as reviewed in, Cisneros et al., 2022). Further clarification of the cause of mitochondrial dysfunction and its impact on Sanfilippo disease progression is needed to establish its value in biomarker development.
Given the ubiquitous expression of mitochondrial proteins throughout the body, peripheral measures will not explicitly measure CNS changes. Despite this, biomarkers of mitochondrial dysfunction have been investigated in AD, where lipids central to mitochondrial metabolism (phosphatidylcholine, lysophosphatidylcholine and acylcarnitine) have been found depleted in plasma up to 2 to 3 years prior to development of mild cognitive impairment or dementia (Mapstone et al., 2014). Identification of primary mitochondrial diseases using blood samples has also been assessed; reduced mitochondrial oxidative phosphorylation and increased anaerobic ATP production (including creatine, lactate and pyruvate) in cells extracted from blood can be detected, as reviewed by Hubens et al. (2022).
Measurements of any ubiquitously expressed proteins (such as those of mitochondrial, autophagic or proteasomal function) in peripheral fluids such as serum/plasma, urine or saliva are unlikely to correlate specifically with CNS disease processes as is required for a Sanfilippo biomarker. To gain CNS specificity, neuronal-and/or astrocyte-derived extracellular vesicles extracted from blood will be necessary; an approach taken by Goetzl et al. (2021), who identified mitochondrial impairment in first episodes of psychosis. A similar approach may prove fruitful in Sanfilippo syndrome, alongside a deeper mechanistic understanding of mitochondrial dysfunction in the disease.

| Neuroinflammation
Widespread neuroinflammation is well documented in the neurodegenerative process of MPS III with gliosis detected repeatedly at post-mortem in patients (Jones et al., 1997;Kurihara et al., 1996;Tamagawa et al., 1985;Viana et al., 2020) and mouse models, where astrocytosis and microgliosis are both well established by 4 weeks of age in MPS IIIA mouse brain (Hassiotis et al., 2014).
Similar changes, albeit at later ages, are reported in the MPS IIIB mouse brain (Wilkinson et al., 2012). Increased levels of proinflammatory cytokines have also been measured in affected mouse brain (Parker et al., 2020;Wilkinson et al., 2012) and patient plasma (Parker et al., 2020). Inhibition of pro-inflammatory cytokine IL-1-driven neuroinflammation reduced astrogliosis and microgliosis and prevented behavioural abnormalities and cognitive decline in MPS IIIA mice (Parker et al., 2020), gov NCT #04018755) arose from these findings. A comprehensive review of how MPS III specific 2-O-sulfated heparan sulfate and various secondary storage products prime and activate the innate inflammatory pathway at Toll-like receptor 4 and NLRP3 inflammasome, respectively, is provided by Mandolfo et al. (2022) following the initial discovery by Ausseil et al. (2008). Given that inflammation is a constantly responsive process in health to fight infection, biomarkers measuring neuroinflammation must target chronic brain-specific changes that are not influenced by episodes of peripheral infection.
Glial fibrillary acidic protein (GFAP) is a key astrocytic cytoskeletal (intermediate filament type III) protein whose expression is increased during astrogliosis (Yang & Wang, 2015), a feature common to many neurodegenerative diseases. With the development of highly sensitive immunoassays such as SIMOA, elevated plasma GFAP levels can now be measured and are useful in predicting pheno-conversion of elderly participants with mild cognitive impairment to dementia (Cicognola et al., 2021), highlighting its prognostic potential.
Plasma GFAP levels have also been found significantly elevated in the neurodegenerative LSD GM2 gangliosidosis (Sandhoff disease) compared to healthy controls. The highest GFAP levels were associated with the most rapidly progressing disease phenotype found in the youngest patients at disease onset (Welford et al., 2022).
Whether GFAP has prognostic potential for Sanfilippo syndrome warrants further investigation.

| Reductions in synaptic function
During normal development, astrocytosis and microgliosis are involved in activity-dependent synaptic pruning, ensuring neuronal networks are built on robust connections, while inactive synapses are selectively pruned away (Lieberman et al., 2019;Schafer & Stevens, 2010;Stevens et al., 2007). Active synapses take up protective factors such as brain-derived neurotrophic factor (BDNF) to further enhance their activity and survival (Lieberman et al., 2019). Given the reductions in BDNF (Villani et al., 2007) and synaptic activity (Dwyer et al., 2017;Pará et al., 2021;Sambri et al., 2017) observed in MPS III mouse tissues, it is possible that over-pruning of underactive neural networks may contribute to progressive neural network disconnection and symptom onset.
Such changes have been observed as early as post-natal day 10 in MPS IIIC mice, where significantly fewer dendritic spines were observed, never reaching wild-type levels and further reduced with age (Pará et al., 2021). At autopsy, reductions in dendritic spines are also observed near mega-neurites (Takashima et al., 1985) and focal dendritic enlargements (Ferrer et al., 1988) in human Sanfilippo syndrome cerebral cortex.
Synapses, distal to the cell body, also rely on the trafficking of vesicles and organelles for their proper maintenance and functioning. Axonal dystrophy, signifying aberrant axonal trafficking, has been observed from 3 weeks of age in the brain of MPS IIIA mice (Beard et al., 2017), and found to contain autophagic organelles and in some cases mitochondria in MPS IIIB mice (Wilkinson et al., 2012). In MPS IIIC mice, decreased synaptic vesicle trafficking to the synapse has been measured, and axon terminals containing multivesicular bodies resembling autophagosomes were found (Pará et al., 2021), both potentially impacting synaptic protein availability for function. Reduction in functional synaptic proteins such as synaptophysin from 10 days of age in MPS IIIB mice (Vitry et al., 2009), vesicle-associated membrane protein by 4 months of age in MPS IIIA and IIIB mice (Wilkinson et al., 2012) and syntaxin-binding protein 1, synapsin-1, calcium/calmodulin-dependent protein kinase II and neuroligin-2 in MPS IIIC mice at 3 months of age (Pará et al., 2021) have all been observed.
Functionally, there is reduced exocytosis of synaptic vesicles in MPS IIIA mouse hippocampal neurons (Sambri et al., 2017) and chromaffin cells (Keating et al., 2012), as well as reduced frequency and amplitude of miniature excitatory and inhibitory post-synaptic currents in MPS IIIC mouse neurons (Pará et al., 2021). Reduced synaptic activity has also been reported in MPS IIIA mouse cells (Dwyer et al., 2017;Sambri et al., 2017). These changes may result directly from reduced synaptic proteins and vesicle availability at the synapse for function.
In AD (Reddy et al., 2005) and frontotemporal dementia (FTD; Ferrer, 1999), various synaptic proteins are reduced in the brain at autopsy and their biomarker potential has been investigated.
When neural-derived exosomes from the blood of AD and FTD patients were assessed for synaptic protein content, reduced levels of synaptophysin, synaptopodin, synaptotagmin-2 and neurogranin were found compared to age-matched healthy controls; synapsin was also reduced in AD (Goetzl et al., 2016). Several of these proteins were decreased 1-10 years prior to AD or FTD diagnosis in the healthy/mild cognitive impairment state of the same AD or FTD patients. Synaptophysin, synaptotagmin and synaptopodin levels also demonstrated a relationship with cognitive performance (Goetzl et al., 2016). Whether these synaptic protein changes can be detected in brain-derived extracellular vesicle extracts from blood in Sanfilippo syndrome warrants further investigation for prognostic, progression and neurocognitive response-to-treatment purposes.

| Cerebral atrophy
Although no gross anatomical differences are seen between the early developing brain of healthy and Sanfilippo patients (Greenwood et al., 1978), the onset of behavioural symptoms is closely associated with cerebral grey matter volume loss, measured by MRI and a key feature of Sanfilippo CNS disease progression (Potegal et al., 2013;Shapiro, King, et al., 2016;. In fact, reduction in grey matter volume has been detected as early as 2.5 years of age in rapidly progressing diseases . Given the limitations of MRI application in the very young, the development of tools and biomarkers of neuroaxonal degeneration in infants and toddlers is warranted to evaluate the prognostic potential of this feature of the disease. 5.6.1 | Novel window to brain denervation-Tool for prognostics/progression Optical coherence tomography (OCT) is a tool widely available to the optometrist, which is capable of imaging the structure of the retina and optic nerve non-invasively. Thinning of the retina's outer nuclear layers correlating with disease progression, and disease prevention by gene therapy in MPS IIIA mice, highlights OCTs' potential in monitoring disease progression and response to treatment (Beard et al., 2020). Findings corroborated in MPS IIIC (Ludwig et al., 2023).
Compliance in children with neurological impairment may be an issue, however, highlighting the need for a blood-based biomarker in the clinic. 5.6.2 | Proteins of neuroaxonal degeneration released into peripheral fluids Neurofilament light (NfL), a neuron-specific cytoskeletal protein, released into CSF and blood following neuroaxonal degeneration is the best biomarker currently available and under investigation in AD, FTD, Amyotrophic lateral sclerosis (ALS), Huntington's disease (HD), acute and progressive inflammatory Multiple sclerosis (MS), traumatic brain injury and cerebrovascular (stroke) (as reviewed by Barro et al., 2020;McColgan & Tabrizi, 2018). Elevated blood NfL could predict disease onset a year or more in advance in AD, familial ALS, HD and MS (as reviewed by Barro et al., 2020;Blennow & Zetterberg, 2018;McColgan & Tabrizi, 2018). The level of serum or CSF NfL at onset of sporadic (non-familial) ALS is strongly prognostic and helps to define those that will progress fast or slowly (Benatar et al., 2022). With regards to neuronopathic LSDs, GM1and GM2-gangliosidosis patient plasma exhibited significantly elevated NfL levels, where the highest levels were observed in rapidly progressing patients (Welford et al., 2022). In Neuronal ceroid lipofusinosis type 2, elevated plasma NfL was measured in all patients compared to controls; and plasma NfL consistently reduced over a 3- year treatment period, demonstrating potential value as a responseto-treatment biomarker (Ru et al., 2019). Finally, in neuronopathic MPS II patients, who store heparan sulfate and dermatan sulfate, elevated NfL in serum has also been found (Bhalla et al., 2020).

Assessment of NfL as a biomarker of neuroaxonal injury in
Sanfilippo has recently been performed in a small cohort of MPS IIIA patients enrolled in a trial of intracerebral AAVrh10-based gene therapy (Clini calTr ials.gov NCT# 03612869). Serum NfL levels at baseline were reduced 24 months post-treatment (Lysogene, 2022 NfL levels in the healthy population have been established from 5 to 95 years of age (Bornhorst et al., 2022;Simrén et al., 2022) and across testing sites (Sejbaek et al., 2020) with this SIMOA kit, aiding its clinical application. For NfL's use prognostically in Sanfilippo, determination of its normal population ranges from birth to 5 years of age in healthy children will be required. The use of NfL as a prognostic biomarker for neuroaxonal injury in other neurodegenerative disorders is an exciting advance likely to broadly promote its cost-effective application and accessibility in the clinic if proven beneficial in the context of Sanfilippo syndrome prognostication. to stratify which pre-symptomatic Sanfilippo infants and toddlers will develop dementia following a newborn diagnosis is essential.

| SUMMARY
Diagnostic biomarkers also provide no accurate measure of disease state/progression or neurocognitive response-to-treatment for Sanfilippo patients. We currently rely on labour-intensive and expensive MRI for cerebral grey matter volume, and behavioural testing for DQ score, neither of which are readily accessible beyond specialist centres nor applicable to the very young. This highlights the need for the development of biomarkers for these contexts of use as well.
The development and implementation of novel biomarkers for any context of use requires that they perform at least as well as current measures in the context, and are a protein known or that makes sense to be affected based on the established neuropathology of disease. For Sanfilippo syndrome, novel prognostic biomarkers must perform pre-symptomatically, otherwise, symptom onset remains the best prognostic tool, while novel progression and neurocognitive response-to-treatment measures need to perform as well as MRI and behavioural testing in symptomatic patients.
Not all proteins detected in the brain are limited to CNS expression.
In neurodegenerative diseases, enhanced prognostic potential has been achieved for proteins expressed across the body when brain-derived exosomes have been extracted from peripheral biofluids (blood/saliva).
Furthermore, with the advances in mass spectrometry, determination of the proteome of brain-derived exosomes may be possible, enabling novel CNS protein changes to be determined in a sample accessible in the periphery of Sanfilippo patients for biomarker development.
Many of the known features (proteins, cellular pathways and gross cerebral pathology) of Sanfilippo neurodegeneration are common to other neurodegenerative diseases, making several proteins with prognostic potential in those diseases potentially relevant to the Sanfilippo context of use. This represents the most logical place to commence biomarker discovery as, if proven effective in the Sanfilippo context, a clinical test applied to multiple diseases will improve its accessibility, application and cost in the laboratory, aiding its clinical implementation and ultimately enabling game-changing outcomes that will advance this field of medicine.

CO N FLI C T O F I NTE R E S T S TATE M E NT
There are no conflicts of interest for any of the authors. K.M.H. has received research funding for the development of therapeutic approaches (enzyme and gene replacement) for Sanfilippo syndrome from Shire Human Genetic Therapies and Lysogene in the past.

PE E R R E V I E W
The peer review history for this article is available at https://www. webof scien ce.com/api/gatew ay/wos/peer-revie w/10.1111/jnc.15891.

DATA AVA I L A B I L I T Y S TAT E M E N T
Data sharing is not applicable to this article because no new data were created or analysed in this Review article.

PE R M I SS I O N TO R E PRO D U CE M ATE R I A L FRO M OTH E R S O U RCE S
Bio RENDER Confirmation of Publication and Licensing Rights: Agreement number RH25CI9Y71 for the Journal of Neurochemistry ( Figure 1).