Exploring human plasma proteomic variations in mucolipidosis type IV.

Introduction

Mucolipidosis type IV (MLIV) is a neurodevelopmental and neurodegenerative disease that results from loss-of-function mutations in the lysosomal transient potential channel mucolipin-1 (TRPML1). Typically, patients exhibit delayed developmental milestones within the first year of life and then plateau in psychomotor function, consistent with that of an 18- to 20-month-old healthy child. 1 As they progress through life, MLIV patients experience increased muscular hypertonicity and deteriorating motor function. 2 , 3 Patients additionally develop a continuous decline in vision due to retinal degeneration and corneal clouding, ultimately leading to blindness. 1 , 3 Our recent review paper provides a qualitative comparison of motor and neuronal functional development between MLIV patients and unaffected individuals. 4 One of the major hallmarks of the disease outside of the nervous system is high blood levels of gastrin that, together with developmental delay and corneal clouding, have been utilized as diagnostic criteria for MLIV. As clinically viable therapies for MLIV are being developed, it is critical to identify biomarkers capable of measuring therapeutic engagement in the CNS, which the diagnostic criteria cannot specifically measure. Thus, there is an essential need to broadly illuminate the effects of MLIV on blood proteomic profiles related to the primary disease pathology in the brain.

Brain pathology in MLIV is consistent with a hypomyelinating leukodystrophy. 4 Notably, these pathologies are accompanied by progressive iron accumulation in the basal ganglia, cerebellar atrophy, 5 accumulation of electron-dense lysosomal inclusions in all cell types, 6 and reactive astrocytes and microglia. 7 , 8 Moreover, brain imaging of MLIV patients demonstrates a shortage of subcortical white matter, hypoplasia and dysgenesis in the corpus callosum, and variable white matter lesions. 5 Over time, patients experience a decrease in subcortical white matter and cerebellar volumes, while cortical gray matter volumes remain relatively intact. 9 A few MLIV patients exhibit milder neurological impairment, which is associated with residual TRPML1 function conferred by certain allelic variants. 2 , 10

MLIV is inherited in an autosomal recessive manner and caused by pathological variants in MCOLN1 , the gene encoding TRPML1. 11 , 12 , 13 TRPML1 regulates multiple lysosomal functions and processes, including Ca 2+ -dependent lysosomal membrane fusion events in endocytosis, exocytosis, and autophagy 14 ; regulation of the transcription factor EB and the CLEAR gene network 15 ; autophagosome membrane assembly 16 ; and chaperon-mediated autophagy. 17

Loss of Mcoln1 expression in mice mimics human MLIV and results in motor and cognitive deficits, accompanied by the characteristic brain pathology hallmarks such as hypoplastic corpus callosum, hypomyelination, glia activation, and lysosomal storage inclusions. 18 , 19 , 20 , 21 Motor deficits in mice become evident at 2 months of age and progress to hindlimb paralysis and premature death at around 7–8 months of age. 21 , 22 Histological assessment of the end-stage brain pathology showed partial loss of cerebellar Purkinje cells, with otherwise largely preserved neuronal populations. 19 , 20

At present, clinical intervention for MLIV is limited to palliative care and management of some of the symptoms; however, CNS-targeted AAV-mediated gene transfer of MCOLN1 was recently shown to restore or prevent motor decline in Mcoln1 −/− mice and ameliorate brain pathology. 22 These preclinical data provide a new hope for developing disease-altering therapies for MLIV. To help advance translational studies of MLIV and future clinical trials, we set out to define a blood plasma proteomic profile of MLIV as a first step toward molecular biomarker discovery.

In this study, we identify a proteomic profile in plasma of MLIV patients and age- and sex-matched healthy controls using a multiplex protein quantification technology, SomaScan. Using DNA-based SOMAmers that are specific to each protein target, the SomaScan assay delivers readouts of more than 7,000 high and low-abundance proteins from a single small quantity serum or plasma sample. We use these data to establish a protein signature of human MLIV, perform pathway analysis, and correlate differentially expressed proteins and pathways with motor function scores in patients to identify protein signatures related to neuro-motor disability and disease progression. We then compare human plasma and MLIV mouse brain proteomes and identify common overlapping protein signatures, providing further insights into disease mechanisms and identifying protein biomarker candidates for future validation.

Results

Typical mucolipidosis type IV patients show differential blood plasma protein profile

We began the current study by asking if we could identify distinctive plasma protein signatures using the SomaScan assay ( Figure 1 A). To do so, we analyzed protein content in plasma samples from 17 MLIV patients ( Table 1 ) and 38 age-matched Biobank controls ( Table S1 ). The assay detected 7,322 proteins and revealed one cluster of upregulated proteins in MLIV patients and a second cluster that was downregulated in MLIV patients ( Figure 1 B; Table S2 ). Differential analysis of protein abundances between MLIV and control groups revealed 1,961 differentially expressed proteins (DEPs; Figure 1 C, adjusted p value < 0.05, and log 2 fold-change >0.29 or log 2 -fold change <−0.22, materials and methods ; Table S3 ). To gain insight into the abundance of DEPs associated with established functions relevant to MLIV, we built an NCBI database scraper to assign MLIV-relevant annotations to each of the DEPs (autophagy, interferon, iron, lipid, muscle, and neuron; see materials and methods ; Table S3 ). Labeling these annotations on the volcano plot revealed that DEPs in all annotation groups were found among both upregulated and downregulated signatures ( Figures 1 C and 1D). Importantly, we found a higher percentage of upregulated DEPs in “Interferon,” “Lysosome,” and “Muscle” annotation groups, whereas more downregulated DEPs were found in “Neuron/Neural” annotation group. Given that these annotations were derived from our understanding of MLIV pathophysiology, these data suggest that the plasma contains protein signatures relevant to the tissue compartments and mechanisms involved in MLIV. Further analysis utilized StringDB to construct a protein-protein interaction network of upregulated and downregulated proteins ( Figures S1 A and S1B).

Plasma proteins in MLIV patients correlate with motor function scores

In MLIV patients, we simultaneously assessed motor function via the Brief Assessment of Motor Function (BAMF) scales and muscle tone via the modified Ashworth scale as previously described. 2 Although the BAMF scale is inherently relative, it is useful for comparing motor function within a single population. Because muscle spasticity is absent in healthy controls, the Ashworth score will be zero for all healthy individuals. With the available data, we next asked if motor function or muscle tone is associated with specific blood protein signatures in MLIV patients. To test this, we calculated Spearman correlation between the 1,961 DEPs in MLIV patients (either up- or downregulated) and key motor function (BAMF) and muscle tone (Ashworth) scores as well as age ( Figures 2 A and 2B; red indicates correlation with age or diminished function; Table S4 ). Among the upregulated DEPs, this analysis revealed a cluster of 212 proteins that “progressively” increased with worsening muscle hypertonicity, Gross Motor and UEGMS scores, and increasing age ( Figure 2 A, cluster 1; Table S5 ). Gene ontology (GO) terms significantly overrepresented in cluster 1 included muscle protein development, TAK1 activation, and extracellular exosome. A second cluster of 254 proteins (cluster 2) was associated with better muscle tone assessed via Ashworth and decreased with age, suggesting an adaptive or possibly “fatigued” upregulation with disease progression. GO terms for cluster 2 included mitochondrion, lysosomal membrane, protein translation, and RNA binding ( Figure 2 A). Among the downregulated DEPs ( Figure 2 B; Table S5 ), this analysis revealed a cluster (cluster 3) of 588 proteins positively correlated with degree of hypertonicity and age, suggesting an early potentially adaptive downregulation that is lost with disease. GO terms enriched in cluster 3 proteins included axon development, signaling, and cell adhesion. Among downregulated DEPs, a second cluster of 139 “progressively” downregulated proteins was correlated with better muscle tone and motor function and decreased with patient age (cluster 4). GO analysis of proteins in this cluster identified functions associated with neuronal development and extracellular matrix (ECM). These findings reveal proteomic signatures consistent with our understanding of MLIV’s natural history, including worsened neuronal function and increasing involvement of metabolic proteins with disease progression.

To better understand how the individual DEPs of MLIV patients correlate with different functional variables, such as gross motor and fine motor function, muscle tone (Ashworth), and age, and whether some of the proteins correlate between these functional variables, we created a Venn diagram ( Figure 2 C; Table S6 ). Proteins were considered to correlate with Ashworth scores if there was significant correlation with at least two individual scores. Importantly, we found that the majority of the significant correlations were with age, indicating age-related “progressive” protein signature of the disease. 2 , 23 Among all DEPs, only GABA type A receptor-associated protein (GABARAP) and S100 calcium-binding protein A6 (S100A6) were significantly correlated with gross motor, fine motor, and Ashworth scores ( Figures 2 C, 2D, and S1 C). GABARAP is a mediator of neurotransmitter receptors that correlates with age and disease severity ( Figure 2 D). S100A6 is a calcium-binding ion channel expressed by pyramidal neurons, 24 and its increase in MLIV is correlated with disease severity. Additionally, low-density lipoprotein (LDL)-receptor-related protein 8 (LRP8) was significantly correlated with both age and improved fine motor function, in a negative and positive manner, respectively. This protein mediates neuroplasticity in the adult brain through interaction with reelin. 25 Similarly, cobalamin-binding intrinsic factor (CBLIF) decreased with age and increased with fine motor function. This protein is critical for the sufficient intestinal absorption of vitamin B 12 in the intestine, 26 which is required for myelination, and its deficiency leads to neurologic pathology. 27 Finally, serine hydroxymethyltransferase (SHMT1) was found to possess a significantly negative correlation with age and increased expression with improved fine motor function ( Figure S1 C). This protein appears to be altered in Smith-Magenis syndrome, reducing glycine availability and promoting N-methyl-D-aspartate receptor (NMDA)-mediated neurotoxicity. 28 All other proteins presented in the Venn diagram ( Figure 2 C) are detailed in Table S6 .

Gene set analysis identifies differentially enriched gene sets in MLIV plasma associated with brain hallmarks of the disease

Having identified up- and downregulated DEPs, some of which correlated with either clinical scores or age ( Figures 1 and 2 ), we next asked if we could identify signatures from existing gene set annotations using the same blood proteomic profiling. To test this, we used gene set variation analysis (GSVA) 29 together with gene sets from the Reactome and Pathway Interaction Database (PID), sourced from MSigDB. Doing so revealed 229 significantly enriched gene sets in MLIV samples (upregulated in MLIV) and 27 significantly enriched gene sets in controls (downregulated in MLIV) ( Figure 3 A). The gene sets enriched in MLIV, included sphingolipid biosynthesis, mitogen-activated protein kinase (MAPK) and AKT signaling, interleukin-1, and autophagy ( Figure 3 A; Table S7 ), all of which are consistent with our prior findings from pathological and proteomic analysis of brains in the MLIV mouse model. 23 , 30 , 31 , 32 Word clouds highlighting the most common terms in gene sets with decreased and increased enrichment were additionally created ( Figures S2 A and S2B). Specific significant gene sets related to known disease-relevant functions were observed in both the increased and decreased enrichment regions ( Figure 3 B). Because GSVA collapses genes into a gene set enrichment score for each subject, we next asked if we could identify pathways that are correlated with muscle tone and motor function scores, similarly to our analysis of proteins in Figure 2 . This revealed a cluster of significantly enriched gene sets that correlated with poorer muscle tone and motor function ( Figure 3 C, “Ashworth–Up” cluster; Tables S8 and S9 )—these included gene sets involved in growth factor and MAPK pathway signaling, synaptic signaling, and autophagy. A second cluster of gene sets correlated with better muscle tone and decreased with age ( Figure 3 C, “Ashworth–Down” cluster) and included gene sets associated with proteinase function, RUNX signaling, and MAPK signaling ( Table S9 ). Among these correlations, 14 gene sets were significantly correlated with at least two gross motor, fine motor, or Ashworth scores ( Figure 3 D; Table S10 ). All 14 of these gene sets were positively correlated with clinical scores ( Figure 3 E), indicating their enrichment scores increase with disease severity. Their functions span immune signaling, mitosis, and cell cycle, reflecting an overall dysregulation of cellular function with clinical severity.

Human plasma proteome shares DEPs with MLIV mouse brain tissue

Because MLIV patient brain samples are not available, our research group recently used mass spectrometry to profile the brain proteome in the mouse model of MLIV. 30 , 33 Armed with these data ( Table S11 ), we next asked if we could identify shared signatures between the mouse MLIV brain and human plasma. Among the 185 brain DEPs and 1,961 plasma DEPs, we found that 45 overlapped between datasets ( Figures 4 A and 4B; Table S12 ). The upregulated overlapping proteins were largely related to lysosomal function, whereas the downregulated proteins were largely related to myelination. Both of these signatures are consistent with our understanding of key disease hallmarks: impaired myelination and modified lysosomal activity. Importantly, GSVA of both datasets using the same Reactome and PID gene sets used earlier ( Figure 2 ) also revealed a shared increased enrichment of gene sets in MLIV human plasma and MLIV mouse brain related to protein processing ( Figures 4 C and 4D; Table S13 ). Collectively, these data emphasize the probable presence of shared features between the brain and the periphery. Because an abundance of proteins are shared between regions and move in the same direction, and many of them are related to the primary pathology found in the MLIV brain, e.g., downregulation of myelination-related proteins and upregulation of lysosomal proteins; our findings support the presence of a plasma protein profile that reports brain pathology. Thus, this work provides a panel of plasma proteins that may be used in future studies for the validation of plasma biomarkers of MLIV.

Discussion

In this study, we comprehensively profiled proteomic differences in blood plasma between MLIV patients and bio-banked controls and correlated differentially expressed proteins with clinical motor function and muscle tone scores.

Motivated by our prior work showing shared cytokine changes in MLIV between human blood and mouse brain, 23 we hypothesized in the current study that we would identify blood signatures associated with the primary neuronal and lysosomal dysregulation found in the MLIV brains. Because data are limited in human MLIV brains, we compared our findings in human blood with mouse MLIV brain proteome, identifying a subset of shared changes between these compartments/species. Importantly, many of the overlapping proteins were involved in lysosomal function ( Figure 4 ). This study is the first to define changes in the blood proteome and identify those that are consistent with the MLIV brain. It thus represents an essential first step in the identification of clinically relevant blood biomarkers of MLIV.

Recently, we reported an AAV-MCOLN1 gene replacement strategy that restored motor function when administered to presymptomatic and symptomatic Mcoln1 −/− mice. Excitingly, these data suggest that MCOLN1 gene therapy may be able to restore motor development in patients rather than simply delaying disease progression. However, the time needed for restored developmental processes to produce a clinically meaningful improvement in function in humans is uncertain and may exceed a clinical trial time frame. As such, there is a critical unmet need for a clinically tractable biomarker to measure TRPML1 activity restoration in the brain. MLIV human brain tissue analysis has been historically restricted by the lack of tissue donation and is currently limited to only three published cases. 7 , 8 , 31

Our findings of multiple lysosomal plasma proteins associated with clinical severity or overlapping between brain and plasma are important because they may directly reflect the central role of MCOLN1 in lysosomal regulation. The protein encoded by MCOLN1 is a lysosomal non-selective cationic channel from the transient receptor potential family, TRPML1 or mucolipin-1. TRPML1 is a versatile regulator of lysosomal function. Many of these functions are attributed to its Ca 2+ transport from lysosomes, including Ca 2+ -dependent lysosomal membrane fusion events in endocytosis, exocytosis, and autophagy 14 ; regulation of the transcription factor EB and the CLEAR gene network 15 ; autophagosome membrane assembly 16 ; and chaperon-mediated autophagy. 17 More recently, TRPML1 was also shown to regulate conjugation of ATG8 to single membranes, or CASM, a non-canonical form of autophagy induced by lysosomal damage, LC3-associated endocytosis, or phagocytosis. 34 In addition to these experimentally established functions of TRPML1 facilitated via transport of Ca 2+ , its role in lysosomal transport and maintaining homeostasis of iron and zinc has also been documented. 35 , 36 , 37

To identify protein signatures associated with clinical manifestations of MLIV, we performed a correlation analysis of protein expression, BAMF (gross motor function, fine motor function, upper extremity gross motor function, articulation, and deglutition), modified Ashworth scores (muscle tone), or age. We found a significant correlation of up- and downregulated proteins with each of these functional variables, as indicated in Figure 2 . Notably, we have identified clusters of up- and downregulated proteins associated with clinical measures attributed to disease severity, shown as clusters 1–4 ( Figure 2 ), and GO analysis identified GO terms and processes enriched in these clusters.

Proteins in cluster 1 represent the group that is upregulated in MLIV as compared to healthy controls and positively correlate with disease severity, i.e., worsening of the fine and gross motor function and muscle rigidity. GO analysis of this protein signature showed enrichment in muscle protein, detoxification, extracellular vesicles, and TAK1-dependent signaling. Proteins in cluster 2 are on average upregulated in the plasma of MLIV patients compared to healthy controls and decrease with worsening muscle tone and age. In contrast, cluster 3 shows the group that, on average, is downregulated in MLIV but increases with age and worsening muscle tone. Thus, both clusters 2 and 3 may demonstrate a “fatigued” response with age. It is interesting to note that protein patterns in clusters 2 and 3 showed that many of the proteins that changed in MLIV plasma with age were associated with increasing rigidity (i.e., worsening muscle tone). This pattern was absent or less pronounced for BAMF measures (i.e., gross/motor function, articulation, and deglutition). These observations are supported by age-correlation data of the neurological assessments in a cross-sectional natural history study, 2 where muscle tone rather than BAMF measures showed a significant correlation with patients' age. The signature in cluster 2 (upregulated and declining with age and rigidity) showed enrichment in mitochondrion, lysosomal membrane, RNA binding, and translation. In contrast, cluster 3 (downregulated but increasing with age and rigidity) was enriched in neurogenesis, axon development, Ca 2+ binding, and cell adhesion. Proteins in cluster 4 are, on average, downregulated in MLIV and declined with worsening muscle tone and fine motor function. GO terms enriched in this group were related to neurogenesis, neuron recognition, and ECM processing.

Interestingly, we found several significantly correlated proteins that overlapped between different clinical measures. For example, higher abundancies of GABARAP and S100A6 were correlated with worse clinical presentation in gross motor function, fine motor function, and muscle tone ( Figures 2 C and 2D). GABARAP, together with GABARAPL1, 2, and 3, form a family of GABA-receptor-associated proteins that, with LC3 proteins, are a part of the ATG8 family, conjugation proteins that are essential for autophagy. Together with the well-established and overlapping functions between LC3 and GABARAPs in auto-phagophore formation, GABARAPs have a broad array of unique functions, both related and unrelated to autophagy that have been recently extensively reviewed. 38 , 39 , 40 , 41 Autophagy-related functions of GABARAPs, distinct from LC3, include autophagy initiation, cargo targeting, autophagosome-lysosomal fusion, TFEB-dependent coordination of lysosomal biogenesis and capacity and selective autophagy, such as mitophagy, ERphagy, as well as autophagic degradation of Golgi and centrosomes. Outside of autophagy, GABARAPs, as implied by their name, are involved in the binding and transport of GABA receptors and, as such, regulation of synaptic transduction and nerve signaling. 42 Interestingly, in a recent study, GABARAPs were shown to be essential for TRPML1-mediated activation of TFEB, providing new evidence of mechanistic relationships between the two proteins. 34

GABARAPs were reported to be altered in several human diseases. Expression of GABARAPs (particularly GABARAP and L1 forms) is inhibited in different forms of cancer, and they are hypothesized to act as tumor suppression factors (reviewed in 38 ). In multiple myeloma, breast cancer, and some other forms, GABARAP expression is used for risk, detection, prognosis, and treatment efficacy evaluation. Changes in GABARAPs expression have also been associated with neurodegenerative diseases, including stiff person syndrome, 43 dementia with Lewy bodies, 44 , 45 epilepsy, 46 and neurodevelopmental syndrome due to chromosomal 17p13.1 microduplication. 47

Our data show that all members of the GABARAP family were significantly upregulated in MLIV plasma ( Table S3 ), and, importantly, two members, GABARAP and GABARAPL1, positively correlated with worsening gross motor, fine motor, and muscle tone (the former) or gross motor and fine motor (the latter) clinical scores in MLIV patients. It is currently unknown whether the increase in GABARAPs in MLIV is due to global inhibition of autophagy or whether it reflects CNS-driven dysregulation of GABA receptor trafficking. The elucidation of the detailed mechanism was beyond the scope of the current study. Yet, the consistent increase of GABARAPs and their significant correlation with disease severity make them strong candidates for follow-up validation as biomarkers.

Another protein that is significantly associated with the worsening of gross and fine motor function and muscle tone identified in the plasma of MLIV patients in our study is S100A6 or calcyclin. It is a Ca 2+ - and Zn 2+ -binding secreted protein with many functions related to cell proliferation and differentiation, cytoskeleton reorganization, and stress response. 48 Changes in secreted S100A6 in body fluids were found to be associated with diseases. Increased serum S100A6 positively correlates with the progression of different forms of cancer and is suggested to be used as a biomarker. 49 , 50 The increase of S100A6 is also linked to fibrosis-related conditions such as liver cirrhosis biliary, chronic renal disease, pulmonary fibrosis, and myocardial infarction. Additionally, high S100A6 is also linked to astrogliosis in Alzheimer disease (AD) and amyotrophic lateral sclerosis (ALS). 24 , 51 The direct mechanistic relationships between loss of TRPML1 and upregulation of S100A6 have not been established, but due to association with the broad spectrum of clinical manifestations in MLIV and its established mechanistic links to Ca 2+ /Zn 2+ regulation and astrocytic pathology in AD and ALS, we conclude that S100A6 can be further evaluated as disease severity markers of MLIV.

We also found that three proteins, LRP8 or ApoER2, CBLIF, and SHMT1, were commonly associated with worse muscle tone, fine motor function, and age in MLIV plasma and, therefore, can also be evaluated as potential disease severity markers of MLIV. Of these three, ApoER2 and CBLIF showed a negative correlation with disease severity and SHMT1 positive. ApoER2 is one of two receptors that bind reelin and facilitate down-stream signaling known for its role in neuron migration in brain development and synaptic modulation. 52 Progressive reduction of CBLIF, a cobalamin-binding intrinsic factor, which is required for cobalamin or vitamin B12 absorption in the ileum, warrants evaluation of MLIV patients for cobalamin deficiency and, if needed, supplementation. SHMT1 is a cytosolic serine hydroxymethyltransferase that has been linked to cancers and cardiometabolic disorders with aging. 53 , 54

To reveal a plasma signature of MLIV related to pathologic alternations in the brain, we have compared protein signatures of DEPs in the plasma of MLIV patients and MLIV mouse brains. Specifically, this comparative analysis with the whole brain tissue 30 , 33 showed concordant protein expression through a downregulation of proteins related to oligodendrocyte development and myelination and an upregulation of lysosomal proteins. Our data identified 21 upregulated and 5 downregulated DEPs, overlapping between the MLIV patient’s plasma and MLIV mouse brain tissue. These proteins present a set of biomarker candidates for MLIV for future validation. Among upregulated, seven (ACTN2, ARL8A, GLB1, GNS, CTSB, ASAH1, and PLD3) were lysosomal proteins; out of five overlapping downregulated DEPs, three (SEMA7A, ENPP6, and CNTN2) were related to oligodendroglia ( Figure 4 B). This pattern highlights the similarity between mouse brain and human plasma protein signatures in MLIV. Other upregulated DEPs that overlapped in MLIV human plasma and mouse brain were mitochondrial proteins (ACYP2, SORD, PARK7, and MPST) or involved in proteasomal protein degradation (PSME2 and PSME1). This study is the first to define changes in the blood proteome and identify those that are consistent with the MLIV model brain. It thus represents an essential first step in the identification of clinically relevant blood biomarkers of MLIV.

We selected the SomaScan platform because of its reported broad dynamic range providing the capacity to analyze blood plasma. 55 Indeed, we were able to identify brain-related signatures in the blood, including myelin, neurons, etc., as well as proteins associated with lysosomal function. We emphasize that this is an early discovery and exploratory study, representing a screen of more than 7,000 proteins across a cohort of just 17 MLIV cases. Further, these signatures have not been validated by an alternative technique in either human or murine tissue. Therefore, any further exploration into the biomarkers identified in this study should proceed only after that target has been sufficiently validated.

These limitations notwithstanding, our findings here, including more than 150 DEPs with a positive correlation with age and 26 proteins with overlapping changes between human blood and mouse brain, represent a strong foundation to inform future biomarker validation studies. We are strongly encouraged by our observation that many of the DEPs in blood plasma are brain proteins and are relevant to MLIV pathogenesis, e.g., myelin. However, another important limitation is that the clinical scores and blood samples were taken as a single snapshot of each patient at different stages rather than a longitudinal analysis. Moreover, the clinical scores are limited in assessing the overall level of neurological dysfunction. 2 Nevertheless, natural history data are broadly limited for MLIV, and this work represents the first study to systematically connect clinical phenotypes with underlying molecular fingerprints of the disease.

In conclusion, this study defines the first blood proteomic profile of MLIV and the first analysis connecting blood proteins to both clinical scores and a brain proteomic signature, albeit in a mouse model. Our analysis identified one subset of markers that are associated with disease severity and a second subset shown to be dysregulated in the brain. As discussed earlier, many of the proteins identified are directly implicated in neuronal and lysosomal function (e.g., GABARAP) as well as myelination, which are several of the central affected functions in MLIV. We posit that proteins associated with lysosomal dysfunction represent a highly promising biomarker panel for MLIV because these are directly connected to the molecular mechanism of loss of TRPML1 function in the disease. Future targeted studies should evaluate how these blood signatures change longitudinally and whether they are affected by forthcoming therapeutic studies.

Materials and methods

Study design and population

Patients with MLIV ( N = 17, F = 10, M = 7) were recruited through the Mucolipidosis Type 4 Foundation. Written informed consent was obtained from legal guardians for participation in our approved natural history and biomarker studies according to protocols approved by the Massachusetts General Hospital Institutional Review Board. Written informed consent or assent was obtained from patients if their neurological capacity allowed. Inclusion criteria for patients included a documented diagnosis of MLIV by (1) clinical or research-based sequencing of MCOLN1 and identification of two pathological MCOLN1 alleles or (2) presence of the expected constellation of clinical symptoms associated with MLIV and documentation of at least one of the following: one pathological MCOLN1 allele, elevated gastrin levels, or a tissue biopsy with evidence of lysosomal inclusions consistent with MLIV. We note that two patients did not have an M COLN 1 genotype as part of their clinical workup. MLIV cases were assigned to either typical ( N = 15) or mild ( N = 2) presentation. Based on our clinical experience, we classified patients as mild if they had obtained independent ambulation at some point in life. All patients were greater than 12 months of age and therefore expected to have obtained independent ambulation. All patients underwent a full examination by a single board-certified pediatric neurologist, and their function was scored with the BAMF scales (gross motor, upper extremity gross motor, fine motor, deglutition, and articulation) and modified Ashworth scale.

With regard to normative data, the BAMF is a series of 10-point ordinal scales developed for rapid description of the highest level of gross, fine, or oral motor function in children. It is independent of age and quality of performance. While the scale was not developed with normative data for comparison of subjects to “normal function for age,” it does provide a standardized scoring system where each item is typically rated on a scale, allowing for comparison of scores within a similar population and tracking changes over time within an individual patient, essentially providing a relative interpretation of motor function. Additionally, the Ashworth scale assigns a numeric value to the severity of muscle spasticity that is absent in health controls. Thus, normative values for the Ashworth should be zero at all ages. Given these characteristics of BAMF and Ashworth, normative data were not measured and are not available. These data are shown in full in Misko et al. 2 Motor scores were measured within 24 h of the collection of patient plasma. Due to limitations associated with patient access and cost, repeat scoring by independent evaluators was not possible. Plasma samples for the control group were obtained from Mass General Brigham Biobank, a biorepository of consented patients’ samples at Mass General Brigham (parent organization of Massachusetts General Hospital and Brigham and Women’s Hospital). The sample cohort includes 38 individuals selected from the 20,270 available plasma samples of patients aged 0–25 years ( Table S1 ). The samples were selected based on four factors: (1) the age of the patient, (2) the Charlson Age-Comorbidity Index, (3) medication history, and (4) the history of medical problems. Samples from subjects with the Charlson Age-Comorbidity Index of 0–2 (corresponding to 98.3%–90.15% 10-year survival rate), taking no medication such as immune stimulants, immune suppressants, hormones/synthetics/modifiers, CNS and cardiovascular medications, and blood products/modifiers/volume expanders and having no medical problems, i.e., no family history of cancer, heart disease, and other chronic diseases, have been selected for this study.

Blood sample collection and processing

Blood was collected in K2EDTA-coated Purple/Lavender top vacutainers (BD 368047). Plasma was isolated within 4 h from collected blood samples by centrifugation at 1,000 × g for 10 min at 4°C. Isolated plasma was stored at −80°C.

SomaScan proteomic assay

Protein quantification was performed using SomaScan assay v.4.1 by SomaLogic (SomaLogic Operating Co., Inc., https://somalogic.com/ ). This is a high-throughput, multiplexed aptamer-based proteomic technology capable of measuring approximately 7,000 unique human protein analytes. Plasma samples from MLIV and control subjects were sent to SomaLogic Inc. on dry ice and analyzed in a single batch. SomaLogic normalized the data for hybridization, intraplate median signal normalization, and plate scaling.

Animal studies and sample collection

Mcoln1 −/− mice were maintained as previously described. 21 Genotyping was performed by Transnetyx using real-time qPCR ( www.transnetyx.com ). The Mcoln1 +/− breeders for this study were obtained by backcrossing onto a C57BL/6J background for more than 10 generations. Experimental cohorts were obtained from Mcoln1 +/− × Mcoln1 +/− mating. Mcoln1 +/+ littermates were used as controls. Five Mcoln1 −/− mice and five Mcoln1 +/+ were used for this analysis. Experiments were performed according to the Institutional and National Institutes of Health guidelines and approved by the Massachusetts General Hospital Institutional Animal Care and Use Committee.

Statistical and multivariate analyses

Differential expression analysis

Differential expression based on relative protein expression results from SomaScan was determined using a two-sided Wilcoxon rank-sum test with the stats package (v4.4.0) in R. The comparison was made between typical MLIV samples and control samples. Prior to calculation, variance stabilizing normalization was conducted with the normalizeVSN function available through the limma package (v3.54.2) in R, which log2 transformed the data. A false discovery rate (FDR)-adjusted p value <0.05 was considered significant. Volcano plots were constructed using ggplot2 (v3.4.1) in R. Log2 fold-change thresholds were determined as the 15th percentile of fold-change values for upregulated and downregulated proteins with an FDR-adjusted p value <0.05. Differential protein expression in the Mcoln1 −/− mouse brain tissue was determined with limma, as is consistent with the source publication. 30 This package was used to fit a linear model to the data and calculate empirical Bayes statistics. An FDR-adjusted p value <0.05 was considered significant as well. For the plasma proteome samples, Wilcoxon rank-sum test offered a more stringent cutoff for differential expression and was thus used.

Gene set variation analysis

To further elucidate functional changes, gene set variation analysis (GSVA) was performed. 29 This analysis was conducted using the GSVA package (v1.44.5) available in R through Bioconductor with the relative protein expression values. Significant differences in gene set enrichment scores between the two groups were determined via gene set permutation analysis, as we have done before. 56 Gene labels were randomly shuffled, and GSVA was recalculated. This procedure was repeated 1,000 times, and p values were computed as the rate that a random difference in gene-set means between groups was greater in magnitude than the true difference. Benjamini and Hochberg’s FDR-adjusted p values <0.05 were considered significant. The gene sets used for this analysis were the human Reactome Database and Pathway Interaction Database (PID) obtained from MSigDB C2 gene sets. A minimum gene set size of 10 was used, leading to a total gene set count of 1,104.

Correlation analysis

Plasma proteins and gene sets were correlated with clinical scores to elucidate potential biomarkers of patient motor function. Protein expression and motor scores for mild MLIV patients ( n = 2) were included in the analysis. Spearman’s correlation was selected due to the expected nonlinear but monotonic response between protein expression and motor function. Correlation statistics were obtained using the “cor” function within the stats package in R. An unadjusted p value < 0.05 was considered significant. Relative expression values were used to correlate individual proteins, and enrichment scores from GSVA were used to correlate gene sets.

Gene ontology analysis

To provide specific functional annotations on correlated clusters of differentially expressed proteins (DEPs), gene ontology (GO) analysis was performed using PANTHER (v8). 57 Protein names were loaded using a csv file for a statistical overrepresentation test. The reference list was set as the 7,322 proteins identified by the SomaScan proteomic assay. Significant terms were determined as those with an FDR-adjusted p value <0.05 following a Fisher’s exact test.

Gene functional annotation

To provide additional functional annotation of the DEPs, a web scraper was built to parse the National Center for Biotechnology Information Gene Database (NCBI) entry for each protein and identify relevant terms. Protein names were connected to the NCBI Entrez ID with the Genome-wide annotation for Human package, org.Hs.eg.db (v3.15.0), available through Bioconductor in R. This tool utilized the HTML parser in the BeautifulSoup4 (v4.12.2) Python package to extract the gene summary information from each gene entry in the database. The script then identified the keywords in the summary.

Protein-protein interaction network

A protein-protein interaction network was produced for the significantly upregulated and downregulated proteins using the StringDB application (v2.0.1) in Cytoscape (v3.10.1). A minimum confidence for each edge of 0.40 was used, ensuring high interaction confidence. Furthermore, a minimum threshold of system involvement for both the nervous system and muscular system of 4.5 was used for upregulated proteins and 3.5 for downregulated proteins. This filtering established the proteins that are believed to be neuronally involved and functionally relevant for motor performance. Singletons were removed from the analysis as well.

Data availability

All data used in this work are included as part of the supplemental Excel file.

Acknowledgments

We thank Mass General Brigham Biobank for providing control samples and health information data and the ML4 foundation, Dr. Rebecca Oberman, and Randy Gold for providing research funding (Y.G.) and help with MLIV blood samples collection. This work was also funded by the UPenn Orphan Disease Million Dollar Bike Run grant (to A.M. 2017D007934) and Georgia W. Woodruff School of Mechanical Engineering Faculty Fellowship at Georgia Tech (L.B.W.). B.T. and L.B.W. were funded in part by the National Institutes of Health under grant number 1R01AG075820. B.T. was funded in part by the Cell and Tissue Engineering Training Grant from the National Institute of General Medical Sciences of the National Institutes of Health under Award Number T32GM145735. The Graphical Abstract was designed using resources from Flaticon.com .

Author contributions

Y.G. and L.B.W. designed and oversaw the execution of the study. A.M. conducted motor function and clinical assessment of motor function scores. V.M.-B. and M.S. coordinated human plasma and mouse brain sample and data collection. B.T. executed computational data analysis. B.T., Y.G., and L.B.W. drafted the manuscript. All authors reviewed the manuscript.

Declaration of interests

Y.G. received research funding from the ML4 Foundation; A.M. received funding from UPenn Orphan Disease Million Dollar Bike Run Grant Program.

Footnotes

Supplemental information can be found online at https://doi.org/10.1016/j.omtm.2025.101479 .

Contributor Information

Yulia Grishchuk, Email: ygrishchuk@partners.org.

Levi B. Wood, Email: levi.wood@me.gatech.edu.

Supplemental information

Document S1. Figures S1 and S2

Data S1. Tables S1–S13

Document S2. Article plus supplemental information

References

1. Altarescu G., Sun M., Moore D.F., Smith J.A., Wiggs E.A., Solomon B.I., Patronas N.J., Frei K.P., Gupta S., Kaneski C.R., et al. The neurogenetics of mucolipidosis type IV. Neurology. 2002;59:306–313. doi: 10.1212/wnl.59.3.306.

2. Misko A.L., Wood L.B., DeBono M., Oberman R., Raas-Rothschild A., Grishchuk Y., Eichler F. Cross-sectional Observations on the Natural History of Mucolipidosis Type IV. Neurol. Genet. 2022;8 doi: 10.1212/NXG.0000000000000662.

3. Chitayat D., Meunier C.M., Hodgkinson K.A., Silver K., Flanders M., Anderson I.J., Little J.M., Whiteman D.A., Carpenter S. Mucolipidosis type IV: clinical manifestations and natural history. Am. J. Med. Genet. 1991;41:313–318. doi: 10.1002/ajmg.1320410310.

4. Misko A., Wood L., Kiselyov K., Slaugenhaupt S., Grishchuk Y. Progress in elucidating pathophysiology of mucolipidosis IV. Neurosci. Lett. 2021;755 doi: 10.1016/j.neulet.2021.135944.

5. Frei K.P., Patronas N.J., Crutchfield K.E., Altarescu G., Schiffmann R. Mucolipidosis type IV: characteristic MRI findings. Neurology. 1998;51:565–569. doi: 10.1212/wnl.51.2.565.

6. Amir N., Zlotogora J., Bach G. Mucolipidosis type IV: clinical spectrum and natural history. Pediatrics. 1987;79:953–959.

7. Folkerth R.D., Alroy J., Lomakina I., Skutelsky E., Raghavan S.S., Kolodny E.H. Mucolipidosis IV: morphology and histochemistry of an autopsy case. J. Neuropathol. Exp. Neurol. 1995;54:154–164.

8. Tellez-Nagel I., Rapin I., Iwamoto T., Johnson A.B., Norton W.T., Nitowsky H. Mucolipidosis IV. Clinical, ultrastructural, histochemical, and chemical studies of a case, including a brain biopsy. Arch. Neurol. 1976;33:828–835. doi: 10.1001/archneur.1976.00500120032005.

9. Schiffmann R., Mayfield J., Swift C., Nestrasil I. Quantitative neuroimaging in mucolipidosis type IV. Mol. Genet. Metabol. 2014;111:147–151. doi: 10.1016/j.ymgme.2013.11.007.

10. Bargal R., Avidan N., Olender T., Ben Asher E., Zeigler M., Raas-Rothschild A., Frumkin A., Ben-Yoseph O., Friedlender Y., Lancet D., Bach G. Mucolipidosis type IV: novel MCOLN1 mutations in Jewish and non-Jewish patients and the frequency of the disease in the Ashkenazi Jewish population. Hum. Mutat. 2001;17:397–402. doi: 10.1002/humu.1115.

11. Bargal R., Avidan N., Ben-Asher E., Olender Z., Zeigler M., Frumkin A., Raas-Rothschild A., Glusman G., Lancet D., Bach G. Identification of the gene causing mucolipidosis type IV. Nat. Genet. 2000;26:118–123. doi: 10.1038/79095.

12. Bassi M.T., Manzoni M., Monti E., Pizzo M.T., Ballabio A., Borsani G. Cloning of the gene encoding a novel integral membrane protein, mucolipidin-and identification of the two major founder mutations causing mucolipidosis type IV. Am. J. Hum. Genet. 2000;67:1110–1120. doi: 10.1016/S0002-9297(07)62941-3.

13. Sun M., Goldin E., Stahl S., Falardeau J.L., Kennedy J.C., Acierno J.S., Jr., Bove C., Kaneski C.R., Nagle J., Bromley M.C., et al. Mucolipidosis type IV is caused by mutations in a gene encoding a novel transient receptor potential channel. Hum. Mol. Genet. 2000;9:2471–2478. doi: 10.1093/hmg/9.17.2471.

14. Colletti G.A., Kiselyov K. Trpml1. Adv. Exp. Med. Biol. 2011;704:209–219. doi: 10.1007/978-94-007-0265-3_11.

15. Di Paola S., Medina D.L. TRPML1-/TFEB-Dependent Regulation of Lysosomal Exocytosis. Methods Mol. Biol. 2019;1925:143–144. doi: 10.1007/978-1-4939-9018-4_12.

16. Scotto Rosato A., Montefusco S., Soldati C., Di Paola S., Capuozzo A., Monfregola J., Polishchuk E., Amabile A., Grimm C., Lombardo A., et al. TRPML1 links lysosomal calcium to autophagosome biogenesis through the activation of the CaMKKbeta/VPS34 pathway. Nat. Commun. 2019;10:5630. doi: 10.1038/s41467-019-13572-w.

17. Venugopal B., Mesires N.T., Kennedy J.C., Curcio-Morelli C., Laplante J.M., Dice J.F., Slaugenhaupt S.A. Chaperone-mediated autophagy is defective in mucolipidosis type IV. J. Cell. Physiol. 2009;219:344–353. doi: 10.1002/jcp.21676.

18. Grishchuk Y., Peña K.A., Coblentz J., King V.E., Humphrey D.M., Wang S.L., Kiselyov K.I., Slaugenhaupt S.A. Impaired myelination and reduced brain ferric iron in the mouse model of mucolipidosis IV. Dis. Model. Mech. 2015;8:1591–1601. doi: 10.1242/dmm.021154.

19. Grishchuk Y., Sri S., Rudinskiy N., Ma W., Stember K.G., Cottle M.W., Sapp E., Difiglia M., Muzikansky A., Betensky R.A., et al. Behavioral deficits, early gliosis, dysmyelination and synaptic dysfunction in a mouse model of mucolipidosis IV. Acta Neuropathol. Commun. 2014;2:133. doi: 10.1186/s40478-014-0133-7.

20. Micsenyi M.C., Dobrenis K., Stephney G., Pickel J., Vanier M.T., Slaugenhaupt S.A., Walkley S.U. Neuropathology of the Mcoln1(-/-) knockout mouse model of mucolipidosis type IV. J. Neuropathol. Exp. Neurol. 2009;68:125–135. doi: 10.1097/NEN.0b013e3181942cf0.

21. Venugopal B., Browning M.F., Curcio-Morelli C., Varro A., Michaud N., Nanthakumar N., Walkley S.U., Pickel J., Slaugenhaupt S.A. Neurologic, gastric, and opthalmologic pathologies in a murine model of mucolipidosis type IV. Am. J. Hum. Genet. 2007;81:1070–1083. doi: 10.1086/521954.

22. De Rosa S., Salani M., Smith S., Sangster M., Miller-Browne V., Wassmer S., Xiao R., Vandenberghe L., Slaugenhaupt S., Misko A., Grishchuk Y. MCOLN1 gene therapy corrects neurologic dysfunction in the mouse model of mucolipidosis IV. Hum. Mol. Genet. 2021;30:908–922. doi: 10.1093/hmg/ddab093.

23. Misko A.L., Weinstock L.D., Sankar S.B., Furness A., Grishchuk Y., Wood L.B. Peripheral Inflammatory Cytokine Signature Mirrors Motor Deficits in Mucolipidosis IV. Cells. 2022;11 doi: 10.3390/cells11030546.

24. Yang F., Ma J., Zhu D., Wang Z., Li Y., He X., Zhang G., Kang X. The Role of S100A6 in Human Diseases: Molecular Mechanisms and Therapeutic Potential. Biomolecules. 2023;13 doi: 10.3390/biom13071139.

25. Telese F., Ma Q., Perez P.M., Notani D., Oh S., Li W., Comoletti D., Ohgi K.A., Taylor H., Rosenfeld M.G. LRP8-Reelin-Regulated Neuronal Enhancer Signature Underlying Learning and Memory Formation. Neuron. 2015;86:696–710. doi: 10.1016/j.neuron.2015.03.033.

26. Andersen C.B.F., Madsen M., Storm T., Moestrup S.K., Andersen G.R. Structural basis for receptor recognition of vitamin-B(12)-intrinsic factor complexes. Nature. 2010;464:445–448. doi: 10.1038/nature08874.

27. Miller A., Korem M., Almog R., Galboiz Y. Vitamin B12, demyelination, remyelination and repair in multiple sclerosis. J. Neurol. Sci. 2005;233:93–97. doi: 10.1016/j.jns.2005.03.009.

28. Elsea S.H., Juyal R.C., Jiralerspong S., Finucane B.M., Pandolfo M., Greenberg F., Baldini A., Stover P., Patel P.I. Haploinsufficiency of cytosolic serine hydroxymethyltransferase in the Smith-Magenis syndrome. Am. J. Hum. Genet. 1995;57:1342–1350.

29. Hanzelmann S., Castelo R., Guinney J. GSVA: gene set variation analysis for microarray and RNA-seq data. BMC Bioinf. 2013;14:7. doi: 10.1186/1471-2105-14-7.

30. Sangster M., Shahriar S., Niziolek Z., Carisi M.C., Lewandowski M., Budnik B., Grishchuk Y. Brain cell type specific proteomics approach to discover pathological mechanisms in the childhood CNS disorder mucolipidosis type IV. Front. Mol. Neurosci. 2023;16 doi: 10.3389/fnmol.2023.1215425.

31. Vardi A., Pri-Or A., Wigoda N., Grishchuk Y., Futerman A.H. Proteomics analysis of a human brain sample from a mucolipidosis type IV patient reveals pathophysiological pathways. Orphanet J. Rare Dis. 2021;16:39. doi: 10.1186/s13023-021-01679-7.

32. Weinstock L.D., Furness A.M., Herron S.S., Smith S.S., Sankar S.B., DeRosa S.G., Gao D., Mepyans M.E., Scotto Rosato A., Medina D.L., et al. Fingolimod phosphate inhibits astrocyte inflammatory activity in mucolipidosis IV. Hum. Mol. Genet. 2018;27:2725–2738. doi: 10.1093/hmg/ddy182.

33. Sangster M.L., Bishop M.M., Yao Y., Feitor J.F., Shahriar S., Miller M.E., Chekuri A.K., Budnik B., Bei F., Grishchuk Y. A blood-brain barrier-penetrant AAV gene therapy improves neurological function in symptomatic mucolipidosis IV mice. Mol. Ther. Methods Clin. Dev. 2024;32 doi: 10.1016/j.omtm.2024.101269.

34. Goodwin J.M., Walkup W.G., 4th, Hooper K., Li T., Kishi-Itakura C., Ng A., Lehmberg T., Jha A., Kommineni S., Fletcher K., et al. GABARAP sequesters the FLCN-FNIP tumor suppressor complex to couple autophagy with lysosomal biogenesis. Sci. Adv. 2021;7 doi: 10.1126/sciadv.abj2485.

35. Dong X.P., Cheng X., Mills E., Delling M., Wang F., Kurz T., Xu H. The type IV mucolipidosis-associated protein TRPML1 is an endolysosomal iron release channel. Nature. 2008;455:992–996. doi: 10.1038/nature07311.

36. Du W., Gu M., Hu M., Pinchi P., Chen W., Ryan M., Nold T., Bannaga A., Xu H. Lysosomal Zn(2+) release triggers rapid, mitochondria-mediated, non-apoptotic cell death in metastatic melanoma. Cell Rep. 2021;37 doi: 10.1016/j.celrep.2021.109848.

37. Kukic I., Lee J.K., Coblentz J., Kelleher S.L., Kiselyov K. Zinc-dependent lysosomal enlargement in TRPML1-deficient cells involves MTF-1 transcription factor and ZnT4 (Slc30a4) transporter. Biochem. J. 2013;451:155–163. doi: 10.1042/BJ20121506.

38. Chen J., Zhao H., Liu M., Chen L. A new perspective on the autophagic and non-autophagic functions of the GABARAP protein family: a potential therapeutic target for human diseases. Mol. Cell. Biochem. 2024;479:1415–1441. doi: 10.1007/s11010-023-04800-5.

39. Nieto-Torres J.L., Leidal A.M., Debnath J., Hansen M. Beyond Autophagy: The Expanding Roles of ATG8 Proteins. Trends Biochem. Sci. 2021;46:673–686. doi: 10.1016/j.tibs.2021.01.004.

40. Schaaf M.B.E., Keulers T.G., Vooijs M.A., Rouschop K.M.A. LC3/GABARAP family proteins: autophagy-(un)related functions. FASEB J. 2016;30:3961–3978. doi: 10.1096/fj.201600698R.

41. Wesch N., Kirkin V., Rogov V.V. Atg8-Family Proteins-Structural Features and Molecular Interactions in Autophagy and Beyond. Cells. 2020;9 doi: 10.3390/cells9092008.

42. Ye J., Zou G., Zhu R., Kong C., Miao C., Zhang M., Li J., Xiong W., Wang C. Structural basis of GABARAP-mediated GABA(A) receptor trafficking and functions on GABAergic synaptic transmission. Nat. Commun. 2021;12:297. doi: 10.1038/s41467-020-20624-z.

43. Dalakas M.C. Stiff person syndrome: advances in pathogenesis and therapeutic interventions. Curr. Treat. Options Neurol. 2009;11:102–110. doi: 10.1007/s11940-009-0013-9.

44. Khundakar A.A., Hanson P.S., Erskine D., Lax N.Z., Roscamp J., Karyka E., Tsefou E., Singh P., Cockell S.J., Gribben A., et al. Analysis of primary visual cortex in dementia with Lewy bodies indicates GABAergic involvement associated with recurrent complex visual hallucinations. Acta Neuropathol. Commun. 2016;4:66. doi: 10.1186/s40478-016-0334-3.

45. Tanji K., Mori F., Kakita A., Takahashi H., Wakabayashi K. Alteration of autophagosomal proteins (LC3, GABARAP and GATE-16) in Lewy body disease. Neurobiol. Dis. 2011;43:690–697. doi: 10.1016/j.nbd.2011.05.022.

46. Shangguan Y., Xu X., Ganbat B., Li Y., Wang W., Yang Y., Lu X., Du C., Tian X., Wang X. CNTNAP4 Impacts Epilepsy Through GABAA Receptors Regulation: Evidence From Temporal Lobe Epilepsy Patients and Mouse Models. Cerebr. Cortex. 2018;28:3491–3504. doi: 10.1093/cercor/bhx215.

47. Komoike Y., Shimojima K., Liang J.S., Fujii H., Maegaki Y., Osawa M., Fujii S., Higashinakagawa T., Yamamoto T. A functional analysis of GABARAP on 17p13.1 by knockdown zebrafish. J. Hum. Genet. 2010;55:155–162. doi: 10.1038/jhg.2010.1.

48. Wang Y., Kang X., Kang X., Yang F. S100A6: molecular function and biomarker role. Biomark. Res. 2023;11:78. doi: 10.1186/s40364-023-00515-3.

49. Wei B.R., Hoover S.B., Ross M.M., Zhou W., Meani F., Edwards J.B., Spehalski E.I., Risinger J.I., Alvord W.G., Quiñones O.A., et al. Serum S100A6 concentration predicts peritoneal tumor burden in mice with epithelial ovarian cancer and is associated with advanced stage in patients. PLoS One. 2009;4 doi: 10.1371/journal.pone.0007670.

50. Zhang J., Zhang K., Jiang X., Zhang J. S100A6 as a potential serum prognostic biomarker and therapeutic target in gastric cancer. Dig. Dis. Sci. 2014;59:2136–2144. doi: 10.1007/s10620-014-3137-z.

51. Lesniak W., Filipek A. S100A6 Protein-Expression and Function in Norm and Pathology. Int. J. Mol. Sci. 2023;24:1341. doi: 10.3390/ijms24021341.

52. Omuro K.C., Gallo C.M., Scrandis L., Ho A., Beffert U. Human APOER2 Isoforms Have Differential Cleavage Events and Synaptic Properties. J. Neurosci. 2022;42:4054–4068. doi: 10.1523/JNEUROSCI.1800-21.2022.

53. Wang Q., Lu K., Du H., Zhang Q., Chen T., Shu Y., Hua Y., Zhu L. Association between cytosolic serine hydroxymethyltransferase (SHMT1) gene polymorphism and cancer risk: a meta-analysis. Biomed. Pharmacother. 2014;68:757–762. doi: 10.1016/j.biopha.2014.08.002.

54. Wu S., Li Y., Zhao X., Shi F.D., Chen J. Multiplex proteomics identifies inflammation-related plasma biomarkers for aging and cardio-metabolic disorders. Clin. Proteom. 2024;21:30. doi: 10.1186/s12014-024-09480-x.

55. Kraemer S., Vaught J.D., Bock C., Gold L., Katilius E., Keeney T.R., Kim N., Saccomano N.A., Wilcox S.K., Zichi D., Sanders G.M. From SOMAmer-based biomarker discovery to diagnostic and clinical applications: a SOMAmer-based, streamlined multiplex proteomic assay. PLoS One. 2011;6 doi: 10.1371/journal.pone.0026332.

56. Galea E., Weinstock L.D., Larramona-Arcas R., Pybus A.F., Giménez-Llort L., Escartin C., Wood L.B. Multi-transcriptomic analysis points to early organelle dysfunction in human astrocytes in Alzheimer's disease. Neurobiol. Dis. 2022;166 doi: 10.1016/j.nbd.2022.105655.

57. Thomas P.D., Ebert D., Muruganujan A., Mushayahama T., Albou L.P., Mi H. PANTHER: Making genome-scale phylogenetics accessible to all. Protein Sci. 2022;31:8–22. doi: 10.1002/pro.4218.

Associated Data

Supplementary Materials

Document S1. Figures S1 and S2

Data S1. Tables S1–S13

Document S2. Article plus supplemental information

Data Availability Statement

All data used in this work are included as part of the supplemental Excel file.