The modifier matrix: emerging roles of ubiquitin-like proteins in Alzheimer’s disease

Abstract

Ubiquitin and ubiquitin-like proteins (UBLs) have emerged as critical regulators of protein homeostasis and cellular signaling, processes that are increasingly recognized as central to the pathogenesis of Alzheimer’s disease (AD). This review explores the expanding roles of UBL modifiers, including SUMO, NEDD8, ISG15, UFM1, and ATG8/ATG12, in the development and progression of AD. We discuss how these post-translational modifications influence key pathological features of AD such as amyloid-beta accumulation and neurofibrillary tangles formation, as well as their impact on neuronal function, proteostasis, and neuroinflammation. Recent advances in our understanding of the enzymatic machinery mediating these modifications, and the interplay between different UBL proteins, offer new insights into the molecular mechanisms underlying AD. Furthermore, we highlight emerging therapeutic strategies targeting UBL pathways, which may provide novel avenues for intervention in AD. By integrating current findings, this review underscores the significance of UBL proteins in AD and identifies future directions for research aimed at unraveling their complex roles in neurodegeneration.

Graphical abstract

Introduction

Alzheimer’s disease (AD) is the most common neurodegenerative disorder, marked by progressive cognitive decline and characteristic neuropathology: extracellular amyloid-β (Aβ) plaques and intracellular neurofibrillary tangles composed of hyperphosphorylated tau [1, 2]. These lesions coincide with widespread synaptic dysfunction, neuronal loss, and chronic neuroinflammation. Despite advances in biomarker-guided diagnosis and disease-modifying candidates, effective and broadly applicable therapies remain limited [1, 2]. A growing consensus frames AD not only as Aβ/tau accumulation, but as a systems-level collapse of proteostasis, the integrated network that folds, traffics, and clears proteins in long-lived neurons under metabolic and inflammatory stress [3, 4].

Post-translational modifications are core regulators of proteostasis. By dynamically adding or removing chemical groups or small proteins, post-translational modifications tune substrate activity, conformation, localization, interaction networks, and turnover in real time [5, 6]. Modifications consisting of small chemical groups such as phosphorylation, acetylation and methylation shape signaling and chromatin states [5, 6]. In AD, aberrant phosphorylation of tau at multiple serine and threonine residues destabilizes its microtubule binding, promotes aggregation into paired helical filaments, and drives neurofibrillary tangle formation. At the same time, aberrant tau acetylation further stabilizes pathogenic tau species [7, 8]. Disrupted clearance, caused by inhibition of ubiquitin-proteasome and autophagy-lysosome systems, impairs the turnover of both Aβ and tau, amplifying proteostatic stress [7, 8].

Ubiquitin is a 76-amino-acid protein universally expressed in eukaryotes and the archetype of covalent protein regulation [9, 10]. Ubiquitylation proceeds typically via a canonical ATP-dependent E1-E2-E3 cascade: the E1 enzyme activates ubiquitin, transfers it to an E2 conjugating enzyme, and E3 ligases then transfer ubiquitin from the E2 to a substrate lysine residue to form mono- or poly-ubiquitin chains [9, 10]. Depending on which lysine of ubiquitin is used for poly-ubiquitin buildup, the chain topology encodes different outcomes. For example, K48-linked chains mostly signal 26S proteasomal degradation, whereas K63-linked chains often mediate trafficking, DNA-damage response, or stress signaling [9–11]. Deubiquitinases edit or remove chains to reset signals [9, 10, 12]. In AD, ubiquitin adducts accumulate on both plaques and tangles, indicating impaired turnover of aggregation-prone proteins [13, 14]. Moreover, reduced proteasome catalytic activity and buildup of ubiquitin-conjugated substrates is evident across multiple models and human brain samples [15]. As such, multiple ubiquitin-pathway enzymes have been implicated in tau and Aβ proteostasis, as reviewed elsewhere [4, 15, 16].

In addition to ubiquitin, an expanding family of ubiquitin-like proteins (UBLs) further increases the complexity and specificity of protein regulation in neurons, highlighting their growing importance for understanding AD. While some UBLs share sequence similarity with ubiquitin, all adopt a similar β-grasp fold and, in most cases, are conjugated via dedicated E1-E2-E3 enzyme cascades using C-terminal glycine motifs (Fig. 1) [17–19]. UBL modification diversifies cellular outcomes by directing substrates to distinct subcellular compartments, remodeling protein–protein interactions, and routing cargo toward degradation or alternative fates [17–19]. UBLs can act as mono-modifications or assemble into polymeric chains with distinct linkage types, can function combinatorially, and are reversible through pathway-specific proteases [17–19]. Most UBLs, including ubiquitin itself, form isopeptide bonds with lysine side chains on substrates, enabling competition among different modifiers for the same lysine residue. Substitution of one UBL for another can profoundly alter signaling outcomes, for example by shifting a substrate’s fate from degradation to stabilization or from immune activation to repression.

Fig. 1.

Ubiquitin and UBL systems: enzymes, chain architectures, and functional outcomes. (A) Ubiquitin and UBL conjugation cascades. Both ubiquitin and ubiquitin-like modifiers (UBLs) are produced as precursors that are processed by specific proteases (e.g., USPs/SENPs/NEDP1/USP18/UFSPs/ATG4) to expose the C-terminal glycine. Activation by an E1 (ATP-dependent adenylation and thioester formation), transfer to an E2, and substrate modification catalyzed by an E3 produce isopeptide (or for ATG8s, amide) conjugates on lysine residues (or phosphatidylethanolamine for ATG8s). Cycles are reversed by the corresponding deconjugating enzymes. (B) Diversity of chain topologies. Ubiquitin and UBL modifications include monoubiquitylation (single ubiquitin or UBL at one site), multi-monoubiquitylation (single ubiquitin or UBL at multiple sites), and polyubiquitylation. Poly-ubiquitin or UBL chains can be homotypic (same linkage throughout) or heterotypic, comprising mixed-linkage and branched architectures. Hybrid chains contain ubiquitin together with other UBLs within the same polymer and are depicted in two colors (green: ubiquitin; light brown: UBL). (C) Enzyme families and representative components. Shown are representative proteases, E1, E2, and E3 enzymes for major modifiers and selected example substrates. CSN: COP9 signalosome, PE: phosphatidylethanolamine. Enzymes involved in URM1 pathway are incompletely defined. (D) Cellular consequences. Ubiquitin and UBL modifications modulate protein stability, subcellular localization, enzymatic activity, and protein-protein interactions through pathways such as proteasomal degradation, autophagy, signaling, and trafficking. Created in BioRender. Yan, T. (2026) https://BioRender.com/n870fn3

Together, ubiquitin and UBLs form a functionally interconnected ‘modifier matrix’ in AD, in which distinct conjugation systems can converge on shared substrates, compete for modification sites, and generate combinatorial signals. This review focuses on: UBL classes small ubiquitin-like modifier (SUMO), neural precursor cell expressed developmentally downregulated 8 (NEDD8), interferon-stimulated gene 15 (ISG15), ubiquitin-fold modifier 1 (UFM1), and autophagy-related protein 8 (ATG8) and ATG12. Converging evidence links dysregulation of these UBL pathways to central AD pathophysiological processes such as Aβ and tau proteostasis, synaptic dysfunction, neuroinflammatory signaling, organelle quality control, as well as the progressive failure of neuronal clearance systems (Fig. 2). We outline the core features of each UBL pathway, summarize evidence for their dysregulation in AD, describe mechanistic links to amyloid and tau pathology, and discuss emerging diagnostic and therapeutic opportunities. Several key questions now shape this emerging field, including whether UBL changes are cell type-specific, which alterations correlate most strongly with AD pathogenesis, and when multiple UBL pathways converge on the same pathogenic process, which pathway acts as the dominant driver versus an amplifier or compensatory response. To provide a concise framework for the detailed sections below, Table 1 summarizes the major UBL pathways discussed in this review, their dysregulation in AD, principal disease-linked mechanisms, and potential translational relevance.

Fig. 2.

Modifier crosstalk links ubiquitin-like pathways to shared pathogenic processes in Alzheimer’s disease. Ubiquitin and UBL pathways exhibit functional crosstalk and converge on major pathogenic processes in AD, including Aβ accumulation, tau pathogenesis, altered autophagy flux, proteasome dysfunction, and neuroinflammation. Rather than acting as isolated modifiers, these pathways form an interconnected modifier network that influences protein fate, proteostasis, cellular stress responses, and disease progression. Different UBLs may contribute to overlapping but non-identical aspects of AD biology, and their combined effects may shape both disease heterogeneity and stage-dependent vulnerability. This convergence highlights the potential of UBL pathways as complementary sources of biomarker information and as candidate therapeutic targets in AD. Created in BioRender. Yan, T. (2026) https://BioRender.com/w3k0xyw

Table 1.

Overview of major ubiquitin-like protein pathways implicated in Alzheimer’s disease

UBL	Dysregulation in AD	Main AD-linked mechanism	Potential translational relevance	References
SUMO	Altered SUMOylation balance in AD, including reduced SUMO2/3 conjugation in some AD models, association of SUMO1 with tau pathology, and genetic association of SUMO-pathway variants with disease risk in select cohorts	Modulates APP processing, tau phosphorylation and aggregation, synaptic plasticity, and glial inflammatory responses; SUMO1 may promote tau pathology, whereas SUMO2/3 may be protective in some contexts	Candidate biomarker at the level of pathway activity or substrate-specific conjugates such as SUMO-tau. Therapeutic potential may require isoform-, substrate-, or cell type-specific modulation rather than broad SUMO inhibition	[39–63]
NEDD8	Nuclear-to-cytoplasmic redistribution in AD neurons; NEDD8 detected in tangles; reduced NEDD8 levels reported in AD brain	Regulates cullin-RING ligase activity, APP processing, neuronal survival, synaptic function, and possibly tau-related pathology	Possible biomarker of altered proteostasis state or subcellular pathway mislocalization NEDDylation pathway is druggable, but global inhibition or activation may be harmful; subcellular redistribution and pathway imbalance may offer more selective translational opportunities	[75–93]
ISG15	Strongly upregulated in AD brain and mouse models; enriched in neurons and microglia, often near plaques and with phospho-tau	Links neuroinflammation to impaired proteasomal and autophagic clearance; promotes tau accumulation and may amplify neuron-glia inflammatory signaling	Promising candidate for AD inflammatory and biofluid biomarker development Therapeutic strategies could include reducing chronic ISGylation or enhancing de-ISGylation while preserving host defense	[92, 121, 124–135]
UFM1	Evidence for pathway imbalance in AD, including increased UFM1-conjugated proteins and reduced soluble UFSP2; pathway changes correlate with pathological tau	Promotes tau aggregation and propagation; also linked to ER homeostasis, translational stress responses, and proteostasis regulation	Substrate-specific UFM1 assays may provide clinically informative biomarker readouts for AD Strong emerging therapeutic interest; UBA5 inhibition or UFSP2 restoration may provide pathway rebalancing strategies, especially in tau-related AD states	[136, 139–143]
ATG8/ATG12	Increased LC3-positive puncta and LC3-II together with p62 accumulation indicate activated but incomplete autophagy flux in AD	Controls autophagosome formation, selective autophagy, and clearance of tau, Aβ, and damaged cellular components; dysfunction contributes to failed cargo clearance and extracellular spread	Useful mainly as a readout of autophagy pathway state and flux failure Therapeutic efforts may be most effective when aimed at restoring completion of autophagy rather than broadly altering conjugation machinery	[151–168]

SUMOylation in AD

Pathway overview

SUMO proteins are conserved across eukaryotes and covalently attached to target proteins through SUMOylation [20, 21]. In mammals, the system comprises SUMO1 through SUMO5 [22]. SUMO2 and SUMO3 share ~ 97% sequence identity, are typically not discriminated and referred to as SUMO2/3. By contrast, SUMO1 shares only ~ 47% identity with SUMO2/3, reflecting functional divergence [23]. SUMOylation of target proteins involves E1 activation by the SUMO1 activating enzyme subunit 1 (SAE1)-SAE2 heterodimer using ATP [24], E2 conjugation, where activated SUMO transfers to Ubc9 [25], and E3 ligation, with protein inhibitor of activated STAT (PIAS) and tripartite motif-containing (TRIM) family ligases [26–28]. Sentrin/SUMO-specific proteases (SENPs) (SENP1-7) are important regulators for maturation of SUMO precursors and mediate the removal of SUMO from substrates [29, 30]. Although only a small fraction of any given substrate is SUMOylated at steady state, SUMOylation regulates numerous processes, including cellular stress responses such as oxidative and proteotoxic stress, which are highly relevant to neuronal vulnerability in AD, as well as nuclear organization and transcriptional control [22, 31–38].

Dysregulation and functional implications in AD

Multiple studies indicate disruptions of the SUMOylation pathway in AD. In aged Tg2576 mice which overexpress human mutant amyloid precursor protein (APP) as well as in the human AD hippocampus, SUMO2/3 conjugation is decreased, while SUMO1 shows little or inconsistent change [39–41]. Across these studies, Ubc9 expression appears relatively preserved in AD brain tissue [40–42]. Mechanistically, it has been found that Aβ directly affects the SUMO machinery. Exposure to Aβ1-42 reduces global SUMO conjugation and lowers levels of SENP3, PIAS1/2, and SAE2 (but not Ubc9) in primary cortical neurons, suggesting impaired activity-dependent SUMOylation relevant to synaptic plasticity [40, 43]. Aβ also induces redistribution of Ubc9 from presynaptic to postsynaptic terminals [44]. Functionally, these Aβ-induced changes in SUMOylation are likely to exacerbate AD pathology. Reduced global SUMO conjugation has been linked to impaired synaptic plasticity and neuronal signaling, processes that are essential for learning and memory. In addition, dysregulated SUMOylation can influence both APP processing and tau modification, thereby contributing to amyloid accumulation and tau hyperphosphorylation. However, these effects appear to be context- and isoform-dependent, as SUMOylation can exert both protective and detrimental roles depending on the substrate and cellular state. Histopathological studies further report SUMO1 colocalization with phosphorylated tau aggregates in AD mouse models [45] and increased immunoreactivity in tau inclusions in AD patient cortex [46]. Genetic evidence supports a contribution of pathway variability to disease risk: single nucleotide polymorphisms in Ubc9 and SAE2 have been associated with sporadic late-onset AD and mild cognitive impairment in select cohorts [47–49].

APP is a direct substrate of the SUMOylation pathway, however, the influence of SUMOylation on Aβ levels remains controversial. In some models, SUMOylation of APP near the β-secretase cleavage site and/or enhancement of SUMO pathway activity has been reported to reduce Aβ production. Increased SUMO2 conjugation improved synaptic and cognitive measures in an APP mouse model without altering plaque pathology [50–52]. In contrast, other studies showed that SUMOylation increased Aβ production by stabilizing beta-secretase 1 (BACE1) protein and enhancing its interaction with substrates. SUMO pathway manipulation in cell models also increased Aβ, including via effects on BACE1 stability or amyloidogenic processing [53, 54]. Overall, available data support isoform- and context-dependent effects, warranting further validation in primary neurons and additional AD models [52].

Another target for SUMOylation studies in AD is tau, and accumulating evidence indicates that this modification contributes directly to tau-driven pathology in AD. Tau is predominantly SUMOylated at lysine 340 (K340) within its microtubule-binding repeat domain [45]. SUMO1 modification at this site is promoted by the E3 ligases PIASxα/PIASxβ and reversed by SENP proteases [55] and promotes tau hyperphosphorylation, which accelerates the buildup of pathological tau. It has been shown that SUMOylation at K340 competes with ubiquitylation at the same site and blocks ubiquitin-mediated proteasomal clearance [46]. Mutating this residue to K340R markedly reduces tau aggregation and phosphorylation, underscoring the pathogenic relevance of this modification. Besides SUMO1, tau is also modified by SUMO2/3, which appears to exert neuroprotective effects. In AD-relevant models, enhancement of SUMO2/3 conjugation has been reported to reduce tau aggregation, phosphorylation, and associated synaptic dysfunction, and to improve cognitive and synaptic outcomes, supporting a protective role of this modification [52, 56, 57]. These findings suggest that selectively enhancing SUMO2/3 signaling may represent a therapeutic strategy, although translation to clinical settings remains to be established [52]. SUMO pathways may also influence tau indirectly via p53 SUMOylation, which has been linked to PP2A inhibition and tau hyperphosphorylation [58]. Furthermore, pharmacological SUMOylation inhibition via ginkgolic acid has shown protective effects in neuronal and rodent AD models [58–60]. However, these compounds are best viewed as mechanistic tools, as their specificity and safety profiles currently limit direct translational application.

The SUMO pathway is also a key regulator of neuroinflammation. Here, SUMO appears to play a largely protective, anti-inflammatory role [61]. In astrocytes, exposure to neurotoxic Aβ peptides induces a state of reactive astrogliosis accompanied by downregulation of SUMO1 conjugation. Restoring astrocytic SUMOylation by overexpressing SUMO1 suppressed this reactive phenotype, indicating that functional SUMO modification is essential for maintaining a homeostatic, non-inflammatory astrocytic state [62]. In microglia, SUMOylation of transcriptional regulators, including nuclear receptors such as PPAR-γ [63], supports repression of pro-inflammatory gene programs. Together, these observations position SUMO as a brake on innate immune activation that may be compromised in AD, with loss of this brake contributing to chronic neuroinflammation.

NEDDylation in AD

Pathway overview

The process of attaching NEDD8 to a target protein, known as NEDDylation, follows the canonical E1-E2-E3 cascade [64]. The E1, NEDD8-activating enzyme (NAE), is a heterodimer of NAE1 (also known as APP-binding protein 1 (APP-BP1)) and ubiquitin like modifier activating enzyme 3 (UBA3) [65]. After ATP-dependent activation, NEDD8 is transferred to one of two E2s, ubiquitin conjugating enzyme E2 M (UBE2M), the mammalian homolog of UBC12, or UBE2F. [66, 67]. These E2s, often assisted by NEDD8 E3 cofactors such as RING box protein 1 (RBX1), RBX2 and defective in cullin Neddylation 1 (DCN1) family proteins, conjugate NEDD8 to a conserved lysine on cullins, the scaffolds of Cullin-RING ligases (CRLs) [68, 69]. Cullin NEDDylation induces CRL activation and enhances substrate ubiquitylation and proteasomal degradation [64, 70]. NEDDylation is a highly dynamic and reversible post-translational modification. The process of removing NEDD8 from its substrates, termed de-NEDDylation, ensures that CRL activity is tightly regulated [64]. This deconjugation is primarily mediated by the COP9 signalosome, a multi-subunit complex [71, 72], and the specific protease DEN1/NEDP1/SENP8 [73]. Together, these systems support the CRL on/off cycle.

Dysregulation and functional implications in AD

Under physiological conditions, NEDD8 is predominantly localized in the nucleus, where it plays crucial roles in maintaining cellular homeostasis by suppressing inappropriate DNA replication, cell cycle re-entry, and apoptosis [74]. However, in AD, this nuclear localization pattern is disrupted [75, 76]. Neurons in the AD hippocampus, including pyramidal and granule cells, exhibit a marked nuclear-to-cytoplasmic shift of NEDD8. Notably, NEDD8 immunoreactivity has been detected in neurofibrillary tangles and other ubiquitinated aggregates in AD brain [77]. Furthermore, NEDD8 protein levels have also been reported to be decreased in AD brain [78], indicating that NEDDylation pathway dysfunction is a characteristic feature of AD pathology.

The most prominent connection between NEDDylation and AD occurs through APP-BP1 [79]. Elevated APP increases APP-BP1, hyperactivates NEDDylation and drives NEDD8 nuclear-to-cytoplasmic translocation in vulnerable hippocampal neurons [75, 76]. In primary neurons, interaction between APP-BP1 and presenilin-1 (PS1) drives the degradation of the PS1 C-terminal fragment and reduces Aβ42. Knockdown of APP-BP1 weakens this interaction, stabilizes the C-terminal fragment of PS1, and elevates intracellular Aβ42 [80, 81]. Conversely, in SH-SY5Y cells, inhibition of the NEDD8-activating enzyme lowers BACE1 and presenilin while upregulating a disintegrin and metalloproteinase 10 (ADAM10), which causes a shift in APP processing toward the non-amyloidogenic route [82]. APP and/or its intracellular domain have been reported as NEDDylation substrates, which inhibits their transcriptional activity [83]. Functionally, APP-driven hyperNEDDylation triggers neuronal cell-cycle re-entry and p53-dependent apoptosis [76, 84, 85]. Taken together, APP-dependent changes in APP-BP1 are consistent with NEDDylation acting upstream of both secretase balance and neuronal survival.

Beyond amyloid, NEDDylation is linked to tau pathology. NEDD8 immunoreactivity in neurofibrillary tangles and cytoplasmic redistribution of NEDD8 in AD neurons suggests this pathway is engaged in tau aggregation and turnover [76, 77]. Notably, direct NEDDylation of tau in cells or in vivo has not been demonstrated. The NEDD8-interacting adaptor negative regulator of ubiquitin like proteins 1 (NUB1) and NUB1L can limit GSK3β-driven tau hyperphosphorylation and aggregation [86], and reroute tau clearance to the autophagy-lysosome pathway [87]. These effects position dysregulated NEDDylation as a contributor to the formation and persistence of tau tangles.

At synapses, NEDDylation of key pre- and postsynaptic proteins (e.g., mGlu7, PSD-95) is required for spine stability, neurotransmitter receptor maintenance, basal transmission, and long-term potentiation [88, 89]. Genetic or pharmacologic inhibition of the NEDD8 E1 enzyme NAE disrupts these processes in Nae1 conditional knockout forebrain excitatory neurons and in hippocampal slice/neuronal models treated with MLN4924 [88, 90, 91]. Conversely, in synaptic insulin resistance, Cullin-7 dependent NEDDylation drives Insulin Receptor Substrate-1 degradation, and NAE inhibition can rescue synaptic plasticity and memory in at-risk aging models, underscoring context-dependence [81, 90]. Although these studies were not performed in AD models per se, synaptic dysfunction is a hallmark of AD, suggesting that impaired NEDDylation of synaptic proteins could contribute to disease-associated deficits in neurotransmission and memory. Collectively, NEDDylation supports proteostasis and synaptic function under physiological conditions but, upon dysregulation, contributes to proteotoxicity. In addition to neuronal roles, emerging evidence suggests that NEDDylation may also influence glial cell function, including pathways linked to neuroinflammation [64, 81]. Given the central role of microglia and astrocytes in AD pathogenesis [92, 93], it will be important to determine whether dysregulated NEDDylation contributes to inflammatory responses in these cell types, although direct evidence in AD models remains limited.

ISGylation in AD

Pathway overview

ISG15 was the first discovered UBL. It comprises two tandem ubiquitin-like β-grasp domains separated by a short linker, with low primary-sequence identity to ubiquitin but similar tertiary folds [94–98]. It is synthesized as an inactive precursor and processed to expose a C-terminal LRLRGG motif required for conjugation [99]. Besides conjugated forms, free ISG15 exists intracellularly, where it mediates non-covalent modulatory interactions, and extracellularly, where it acts as a cytokine [100–102]. ISG15 is robustly induced by type I interferons (IFNs) through an IFN-stimulated response element containing promoter, which is activated primarily via the interferon-stimulated gene factor 3 (ISGF3) complex (signal transducer and activator of transcription 1 (STAT1), STAT2, interferon regulatory factor 9 (IRF9)), with additional IRF-family contributions depending on cell context [103–106]. It is also induced by additional inflammatory and stress cues [97, 107–112].

ISGylation of target proteins also involves a three-step cascade: the E1 UBE1L/UBA7 activates ISG15 to form a thioester with its catalytic cysteine [96], the E2 UbcH8/UBE2L6 accepts ISG15 by transthiolation [113], and E3 ligases including HERC5 (HERC6 in mouse), TRIM25, and ARIH1/HHARI ligate ISG15 to substrates. The general understanding is that HERC5 acts broadly while TRIM25/HHARI more selective [114–116]. HERC5 is IFN-induced and associates with ribosomes, enabling co-translational ISGylation. The principal de-ISGylase ubiquitin-specific peptidase 18 (USP18), which itself is strongly induced by IFNs and other inflammatory/genotoxic cues, removes ISG15 from modified proteins [117, 118]. Beyond de-ISGylation, USP18 also dampens type I IFN signaling by binding interferon alpha and beta receptor subunit 2 (IFNAR2) via STAT2, displacing Janus kinase 1 (JAK1) and limiting downstream ISG induction [119]. In human cells, this negative-feedback function is further stabilized by a direct ISG15-USP18 interaction [120].

Dysregulation and functional implications in AD

Under physiological conditions, ISG15 is expressed at low levels in normal cells and tissues, but in AD, it is consistently upregulated across human brains and experimental models [121]. Immunohistochemistry reveals increased ISG15 protein in AD cortex compared to age-matched controls. In 5xFAD mice, both hippocampal ISG15 mRNA and protein are elevated, and ISG15 colocalizes with microglia surrounding amyloid plaques [121]. Similarly, ISG15 is upregulated across hippocampal subregions in triple-transgenic AD (3xTg-AD) mice, where it localizes to neurons and microglia [121]. Co-immunostaining shows that increased neuronal ISG15 coincides with elevated phospho-tau (pT205), suggesting a link to tau pathology. Beyond AD, aberrant ISG15 elevation has been observed in other neurodegenerative conditions such as Amyotrophic Lateral Sclerosis [122, 123]. Together, these findings indicate that ISG15 upregulation is a shared feature of neurodegeneration, although whether it plays a causal role remains to be determined.

Mechanistically, ISG15 links AD pathology to a sustained neuroinflammatory program centered on type I IFN signaling [92, 121, 124]. In AD, a persistent IFN-stimulated gene signature is evident in human postmortem cortex and in mouse models, with ISG15 among the most upregulated transcripts and proteins [92, 121]. This IFN activation likely arises when the innate immune system interprets endogenous danger signals, including amyloid aggregates and tau aggregates, as foreign [92, 124–126]. Microglia exposed to such aggregates produce type I IFNs that induce ISG15 and its conjugation machinery in neurons and glia, establishing an inflammatory milieu that can appear before overt plaque or tangle burden [92, 121, 127].

Within this inflammatory environment, ISG15 functions as both effector and messenger. Stressed neurons upregulate ISGylation and accumulate free ISG15 that can be released into the extracellular space [128–131]. Free ISG15 engages microglial integrins such as CD11b (Mac-1), driving pro-inflammatory cytokine and chemokine production. This interaction also promotes process extension toward plaques and enrichment of reactive microglial states around dystrophic neurites [128–130]. Neuronal ISG15 can also propagate neuron-to-glia signaling (including via extracellular vesicles), and astrocytes in regions of injury show spatially restricted ISG15 induction [93, 122, 128]. Together, neuron-to-glia ISG15 signaling and glial ISG induction help explain why IFN signatures correlate with pathology severity and can precede clinical decline.

In neurons, ISG15 induction correlates with phospho-tau and can be driven by tau overexpression itself, placing tau upstream of ISG15 [121]. Once elevated, ISG15 impairs two major proteostatic pathways. First, ISG15 perturbs the ubiquitin-proteasome system by competing with ubiquitin for E2 and E3 ligases and by generating non-degradative mixed ubiquitin-ISG15 chains that do not efficiently signal proteasomal degradation [132–135]. These effects reduce effective polyubiquitylation and stabilize aggregation-prone substrates, including tau. Second, ISG15 disrupts autophagic clearance by binding and inhibiting histone deacetylase 6 (HDAC6), a deacetylase critical for aggresome formation and autophagosome-lysosome fusion [121]. This leads to impaired autophagic flux, and reduced tau degradation. Experimental knockdown of ISG15 restores autophagy and lowers tau levels, whereas ISG15 overexpression impairs clearance and increases tau burden [121]. Overall, prolonged ISG15 elevation in neurons during AD correlates with reduced proteasomal and autophagic clearance.

Collectively, available data support a model of a potentially self-amplifying ISG15-tau circuit. However, the temporal order, cell-type specificity and necessity in human AD remain to be fully established. This feedforward loop might be interrupted by therapies that modulate ISG15 induction/ISGylation or relieve HDAC6-dependent autophagy deficits while preserving essential immune pathways.

UFMylation in AD

Pathway overview

UFMylation is an understudied UBL pathway with emerging roles in endoplasmic reticulum (ER) homeostasis, autophagy, DNA-damage responses, and immune signaling [136]. UFM1 shares only ~ 15% sequence identity with ubiquitin yet adopts a similar β-grasp fold [137]. Analogous to ubiquitylation, UFM1 is conjugated to substrate lysines by a dedicated cascade consisting of the E1 enzyme ubiquitin like modifier activating enzyme 5 (UBA5), E2 enzyme ubiquitin like modifier conjugating enzyme 1 (UFC1), and an ER-anchored E3 ligase complex comprising UFM1 specific ligase 1 (UFL1) with the scaffold DDRGK domain containing 1 (DDRGK1) and adaptor CDK5 regulatory subunit-associated protein 3 (CDK5RAP3). UFM1-specific protease 1 (UFSP1) processes pro-UFM1 to its mature form by exposing the C-terminal Val-Gly motif required for conjugation, whereas UFSP2 removes UFM1 from substrates [138]. UFMylation is essential for brain development as biallelic loss-of-function in core components causes severe neurodevelopmental disorders [136].

Dysregulation and functional implications in AD

Biochemical and transcriptomic data indicate a profound UFMylation pathway imbalance in human AD. Single-nucleus transcriptomics shows reduced expression of multiple UFM1-pathway components in excitatory neurons, while biochemical analyses reveal increased conjugated UFM1 and decreased soluble UFSP2 in AD cortex [139, 140]. Across cohorts, UFM1 abundance correlates positively with pathological tau species (insoluble total tau, soluble p-tau, insoluble p-tau) but not with total tau, supporting a shift toward hyper-UFMylation that tracks with tau proteotoxicity [139]. A pair of recent studies nominate UFMylation as a key positive regulator of tau aggregation and spread. Using a genome-wide CRISPR interference (CRISPRi) screen in induced pluripotent stem cell-derived neurons, Samelson et al. found that repressing core UFMylation genes (UFM1, UFL1, UBA5, and UFC1) reduced tau oligomers [141]. Parra Bravo et al. validated this in an independent 4R-tau cell model and showed that knocking down the E1 enzyme UBA5 curtailed tau propagation in PS19 mice [142]. Together, these data indicate that the UFM1 cascade promotes tau pathology in vitro and in vivo. Although the molecular mechanisms are unclear, it has been proposed that UFM1 may monoUFMylate tau, potentially increasing its secretion, as seen with α-synuclein in Parkinson's disease [143]. Beyond its role in tau aggregation, UFM1 is linked to several pathways impacted in AD, including ER proteostasis and unfolded protein response regulation, autophagy processes, DNA damage responses, and immune signaling. We have recently provided a comprehensive summary of these connections, including mechanisms, substrates, and therapeutic implications [136].

ATG8/ATG12 in AD

Pathway overview

Autophagosome formation, critical for autophagy, relies on the sequential action of ATG8 and ATG12, two UBLs [144, 145]. First, ATG12 is conjugated to its only substrate, ATG5, in a process requiring the E1-like enzyme ATG7 and the E2-like enzyme ATG10 [144, 145]. This ATG12-ATG5 conjugate then assembles with ATG16L1 to form a large multimeric complex that functions as a novel E3-like enzyme [146]. However, instead of catalyzing ubiquitylation of proteins, its role is to facilitate the modification catalyzed by the second UBL system. ATG8 family proteins, comprised of microtubule-associated protein 1 light chain 3 (LC3) (LC3A/B/C) and gamma-aminobutyric acid receptor-associated protein (GABARAP) (GABARAP, GABARAPL1, GABARAPL2) subfamilies, are proteolytically primed by ATG4s to expose a C-terminal glycine [144, 147]. ATG8 proteins also use ATG7 for activation but a different E2-like enzyme, ATG3, for conjugation. Remarkably, ATG8 UBLs are not attached to another protein but to a lipid, phosphatidylethanolamine, generating membrane-anchored LC3-II/GABARAP-II that drives phagophore elongation and maturation. Lipidated ATG8 proteins aid membrane tethering and fusion. They act as docking stations for selective autophagy receptors such as p62/SQSTM1, optineurin (OPTN), and others, thereby bridging ubiquitinated cargo to the autophagosomal membrane for degradation [148–150]. Together, ATG8 and ATG12 coordinate the membrane dynamics and substrate specificity necessary for functional autophagosome biogenesis.

Dysregulation and functional implications in AD

Autophagy in AD is dysregulated, with increased autophagosome formation but incomplete degradation [151, 152]. In AD brain, neurons show elevated LC3-II (lipidated ATG8) and increased LC3-positive puncta together with p62/SQSTM1 accumulation, indicating deficient substrate clearance [153]. Transcriptomic analysis reveals age- and context-dependent increases in ATG16L1 and GABARAP family transcripts in hippocampal neurons and cellular AD models, suggesting compensatory upregulation of the conjugation machinery in response to proteostatic stress [154]. Converging pathology, vesicle accumulation and lysosomal dysfunction, further amplifies the blockage of execution and tracks with neuritic dystrophy and synaptic failure [153].

These defects have direct consequences for tau and Aβ handling. For tau, selective autophagy requires lipidated ATG8s to engage with autophagy receptors (p62/SQSTM1, OPTN, nuclear dot protein 52 (NDP52), neighbor of BRCA1 gene 1 (NBR1)) that bind phosphorylated/aggregated species and target them for degradation [155–158]. When the flux is compromised, tau progressively accumulates in a manner that is at least partially dependent on the co-chaperone BCL2-associated athanogene 3 (BAG3). While BAG3 upregulation enhances degradation, BAG3 depletion permits tau buildup [159, 160]. LC3-associated endocytosis further facilitates the internalization and degradation of extracellular tau aggregates in microglia [161]. For Aβ, impaired autophagosome-lysosome turnover increases intra-neuritic Aβ production (autophagic membranes concentrate APP and secretases). At the same time, peptide clearance is decreased, promoting plaque-proximal dystrophy [162–164]. Beyond neurons, microglial ATG8-dependent autophagy supports plaque phagocytosis. Its dysfunction reduces Aβ clearance and contributes to sustained neuroinflammation [163, 165, 166]. Finally, secretory consequences of a stalled pathway propagate cell-to-cell transmission as autophagy suppression increases Aβ and tau in extracellular vesicles [167]. Conversely, autophagy induction lowers exosomal cargo, linking ATG8 to decreased extracellular spread [168].

Other UBLs with limited evidence in AD

Unlike the UBLs discussed above, FAT10, also known as ubiquitin D, and ubiquitin-related modifier 1 (URM1) remain less well studied in AD. Nevertheless, they intersect with inflammatory, mitochondrial, translational, and condensate-related pathways that are increasingly implicated in neurodegeneration. We therefore summarize these pathways briefly.

FAT10 is a di-UBL induced by pro-inflammatory cytokines [169, 170], its conjugation uses UBA6 (E1) and USE1/UBE2Z (E2), with emerging, context-dependent E3s [171, 172]. While consistent FAT10 upregulation in AD cortex has not been established, the shared E1 UBA6 is significantly increased in AD brain [173], implying potential pathway flux. Notably, Parkin, the Parkinson’s disease associated E3 ligase involved in mitochondrial quality control, functions as a FAT10 E3, undergoing auto-FAT10ylation and FAT10ylating Mitofusin-2, which suppresses Parkin activation and can impair mitochondrial autophagy under stress [174]. Additional E3 ligases such as ZNF598 may mediate FAT10 conjugation in a USE1-independent manner [175]. While the full impact of FAT10ylation in AD pathogenesis remains unclear, its interface with neuroinflammatory and mitochondrial pathways, both critical in AD, warrants further study.

URM1 is a conserved UBL and sulfur-carrier. It is activated by UBA4 and supports tRNA U34 thiolation, translation fidelity, and oxidative-stress defense [176–179]. Its E2-E3 counterparts are not defined [180, 181]. In yeast, stress-induced cytosolic acidification triggers URM1 self-association and binding to Uba4 and target proteins [182]. URMylation drives stress-sensitive proteins into stress granules and nuclear condensates. URM1 behaves like a reversible molecular adhesive. Its loss causes condensate defects and reduces stress resilience. There is no strong evidence of URM1 dysregulation in human AD brain. Any role in AD might be inferred from stress, translation, and condensate formation. Given the prominence of RNA/protein condensates in neurodegeneration [183], it will be important to test URM1 function separately in neurons, astrocytes, and microglia within tau- and amyloid-based models to see whether URM1 influences condensate dynamics, translation, and stress resilience.

Biomarker potential and challenges of UBLs

During AD pathology, there are distinct changes in several UBL pathways that could be exploited as biomarkers. For instance, total UBL levels of SUMO, NEDD8, ISG15, UFM1, and ATG8/LC3-GABARAP can be monitored to index the overall pathway load. While substrate-specific or chain-resolved measurements remain mostly aspirational at this point, UBL-conjugated sentinel substrates such as SUMO-tau, could be measured by enzyme-linked immunosorbent assay to capture pathway engagement. Direct measurement of these UBLs and conjugates in cerebrospinal fluid (CSF) or plasma would be preferred if technically feasible. Notably, pilot data already show plasma SUMO1 elevation in AD and detectable ISG15 in CSF with levels tracking neuroinflammatory disease severity in HIV-associated neurocognitive disorder [184, 185], supporting the feasibility of direct biofluid assays. For low-abundance targets, enrichment of extracellular vesicles from CSF or plasma can enhance detection sensitivity and improve the brain specificity of biomarker measurements [186]. A key practical challenge is that, although these UBL conjugates are covalent and chemically stable, they can be edited in biospecimens by active enzymes. Therefore, strict control of sample handling prior to analysis is needed to allow reliable comparison between studies. Finally, integrating UBL readouts with established AD biomarkers (Aβ42/40, p-tau217, p-tau231, neurofilament light (NfL), and Glial fibrillary acidic protein (GFAP)) in longitudinal cohorts, with confirmation across orthogonal assay platforms (e.g., immunoassay and mass spectrometry), may enable biologically defined patient subtypes that help stratify patients for targeted interventions [187]. A major next step will be to determine whether UBL profiles can define biologically meaningful AD subtypes or endotypes that improve mechanistic stratification beyond existing biomarker frameworks.

The identification of specific UBL modifications on substrates is technically challenging and remains a major bottleneck. Typically, samples are proteolytically cleaved to identify remnants of conjugation by mass spec. However, while enabling identification of primary conjugation sites, proteolytic digestion causes the loss of critical information about UBL chain architecture such as linkage specificity, chain length, and branching, thereby preventing comprehensive analysis of UBL conjugate topology across all UBL families [188]. In addition, ubiquitin, NEDD8, and ISG15 all leave identical di-glycine (K-ε-GG) remnants after trypsin digestion, making it difficult to distinguish them by standard mass spectrometry. Trypsinization of a second group of UBLs creates large, sometimes branched adducts, which are significantly more difficult to identify by mass spec. For a third group of UBLs such as UFM1 or SUMO, although distinct remnants exist, sensitive and specific enrichment methods are still underdeveloped. For UFM1, a first lysine-ε-Val-Gly di-glycine (K-ε-VG) remnant antibody has been described for enrichment of UFMylated peptides [189], but access is limited and broader validation and standardized workflows are still lacking. These technical limitations have also hindered efforts to determine which UBL modifications occur on the same lysine residues of shared substrates, such as tau or trafficking factors.

Several UBLs, including SUMO, NEDD8, ISG15, UFM1, and ubiquitin itself, compete for conjugation at the same lysine residues on shared substrates. In addition, they can form hybrid chains, where different modifiers co-assemble on the same substrate, which can change the substrate fate and signaling [190]. The best characterized mixed chains are SUMO-ubiquitin hybrids [191, 192], where polySUMOylated substrates recruit SUMO targeted ubiquitin ligases that attach K48- and K63- ubiquitin, switching the fate of targets between proteasomal degradation and signaling [190, 193, 194]. NEDD8- or ISG15-ubiquitin hybrids have also been observed [135, 190, 195–197]. Because hybrids generally seem stress-induced, their abundance and linkage topologies could shift in AD. While SUMO-NEDD8 hybrid chains have been reported [198], their prevalence and function in AD remain unclear and warrant exploratory study. An important unresolved question is whether hybrid UBL chains are enriched in AD brain or biofluids and whether they might serve as sensitive readouts of pathway convergence under disease-associated stress. A systematic profiling of hybrid chains in AD is technically challenging and has not been performed. Hybrids might be good biomarkers of pathway convergence, but robust measurements require chain-specific mass spectrometry, neo-epitope or linkage-specific antibodies. This should be paired with functional assessment UBL modifications in neuronal and glial cells to determine whether a certain modification acts as driver, modulator, or sentinel readout.

Importantly, different UBLs may provide complementary biomarker information. For example, ISG15 may reflect neuroinflammatory activation, whereas SUMO or NEDD8 may report on proteostatic or synaptic dysfunction. A combinatorial UBL signature may therefore better capture disease heterogeneity and progression than individual markers.

Therapeutic targeting of UBL Pathways: opportunities and challenges

The growing understanding of how UBL pathways are dysregulated in AD opens up a new landscape of therapeutic possibilities that move beyond direct targeting of Aβ and tau aggregates. Instead of trying to clear the pathological proteins after they have already accumulated, these strategies aim to correct the underlying failures in the cellular regulatory machinery that allowed them to accumulate in the first place. However, the complexity of the UBL modifier matrix means that the most promising approaches will likely require selective and context-dependent modulation of these pathways rather than simple, broad-spectrum inhibition or activation.

Multiple SUMOylation inhibitors, deSUMOylation inhibitors, and one SUMOylation activator have been identified [199], offering a robust pharmacological toolkit for modulating SUMO pathways. However, the therapeutic potential of SUMO in AD is highly context- and isoform-dependent. Inhibiting SUMO, especially SUMO1-tau K340, can reduce tau hyperphosphorylation and proteotoxicity [46, 55, 59]. Preclinical studies show that compounds like ginkgolic acid can decrease tau hyperphosphorylation and cognitive impairment in Aβ-induced AD rat models [59, 60]. Conversely, APP SUMOylation near K587/K595 may suppress amyloidogenic processing, and enhancement of SUMO2 improves behavior in APP mice, suggesting a neuroprotective role [50–52]. Because broad SUMO inhibition can have variable effects on tau, APP, and inflammation, it may exacerbate neuroinflammation if glial cell SUMOylation is disrupted [56, 57]. Future strategies should therefore focus on substrate-specific or cell-type-targeted inhibitors, as well as isoform-specific modulation to distinguish between SUMO1 and SUMO2/3 signaling.

The NEDD8 pathway is also druggable and several different NEDDylation inhibitors have been identified [200]. The NAE inhibitor MLN4924 (pevonedistat), developed for oncology, is the best-characterized tool for pathway modulation. Benefits of MLN4924 include a shift of APP processing toward the non-amyloidogenic route [82]. In addition, synaptic NEDDylation blockade (e.g., at Cullin-7) can preserve insulin signaling and rescue plasticity under high amyloid burden [90]. Potential liabilities are that global suppression disables proteostasis and synaptic function, may impair Parkin-dependent mitochondrial quality control and other clearance pathways [75]. Notably, NEDD8 protein levels are reduced in AD brain [78], arguing that further global inhibition could be counterproductive. Overall, efficacy and safety are cell-type and disease stage-dependent. Given the consistent nuclear-to-cytoplasmic NEDD8 redistribution in AD [75, 76], normalizing subcellular NEDD8 distribution merits exploration as a more targeted approach.

Compared to SUMO and NEDD8, ISG15 inhibition has a more convergent therapeutic potential. Chronic ISGylation correlates with impaired proteasome and autophagy flux, HDAC6 inhibition, and tau accumulation, while ISG15 knockdown improves pathology and cognition in preclinical models [121]. Although ISG15 has antiviral roles, current data indicate that sustained ISG15 induction in AD is deleterious [92, 131]. Despite its therapeutic potential, drug-like ISGylation inhibitors remain scarce. No selective small-molecule inhibitors targeting the ISG15 E1, E2, or E3 ligases have been validated in cells. Existing chemotypes primarily inhibit USP18, the de-ISGylase, which increases ISGylation [201, 202]. Therefore, alternative strategies such as antisense oligonucleotides or siRNA to ISGylation, with cell-type-targeted delivery to neurons or microglia, may be needed. In parallel, pharmacologically enhancing de-ISGylase activities may offer another route to restoring balance without compromising the antiviral defense.

UFM1-conjugated proteins are increased in AD cortex, and soluble UFSP2, the major de-UFMylase, is reduced and negatively correlated with UFM1 levels, suggesting a pathogenic shift toward hyper-UFMylation [139]. Functional studies show that knockdown of the UFM1 E1 enzyme UBA5 reduces tau aggregation and propagation in both cultured neurons and PS19 tauopathy mice [142], providing strong proof-of-concept for therapeutic benefit. Emerging small-molecule UBA5 inhibitors offer an entry point to modulate the pathway [203, 204]. However, because UFMylation is essential for ER homeostasis and translational quality control [136], specific inhibition of UFM1 accumulation at AD-associated substrates might be needed. In addition, restoring UFSP2 activity may offer a more physiological rebalancing strategy. Enhancing UFSP2 via stabilizers or activators could relieve conjugation pressure without full inhibition. Given that the UFM1-AD connection has only recently come into focus, next steps should emphasize tool and model development, such as selective UFSP2 modulators, to rigorously define therapeutic windows for UFM1-pathway modulation in AD. In contrast to UBLs with context-dependent or protective roles, current evidence positions UFMylation as a predominantly pro-pathogenic pathway in AD, particularly in the context of tau aggregation and propagation. Within the broader modifier matrix, UFM1 may therefore represent a node that actively drives proteotoxicity, making it an attractive candidate for targeted inhibition.

While ATG8/LC3–GABARAPs are not conjugated to protein substrates, they undergo UBL-like lipidation critical for autophagosome maturation. Although they are commonly used as readouts of autophagic flux (e.g., LC3-II), direct pharmacologic modulation of ATG8ylation remains exploratory. Both ATG8 and ATG12 conjugation systems are essential for a wide array of processes, including aggregate clearance, organelle quality control, and cellular stress responses. As a result, broadly interfering with these pathways could have unintended and potentially harmful consequences, emphasizing the need for caution when considering therapeutic strategies that target this axis.

Among current candidates, ISG15 represents a particularly attractive target due to its convergent effects on neuroinflammation and proteostasis, whereas SUMO and NEDD8 pathways may require more nuanced, substrate- or isoform-specific modulation. However, a key implication of the modifier matrix concept is that therapeutic strategies should aim to rebalance, rather than uniformly inhibit or activate, UBL pathways. Because different modifiers exert opposing effects, broad pathway inhibition may produce unintended consequences. Instead, selective and context-dependent modulation of specific UBLs, substrates, or cell types may be required to restore proteostatic balance in AD. Translating these approaches to the clinic presents several challenges. UBL pathways are highly conserved and regulate essential cellular processes, raising concerns about toxicity with systemic modulation. In addition, achieving brain penetration and cell-type-specific targeting, particularly in neurons versus glial cells, remains a major hurdle. The timing of intervention may also be critical, as early modulation of proteostasis and inflammation may be beneficial, whereas late-stage intervention could disrupt compensatory mechanisms. Addressing these challenges will require the development of targeted delivery systems and biomarkers to guide patient stratification and therapeutic window selection. A central translational challenge will be to pinpoint which specific components within UBL pathways can be targeted most safely and effectively for partial therapeutic rebalancing, while preserving essential physiological functions.

Conclusions and future perspectives

Rather than acting as isolated pathways, UBL systems form an interconnected regulatory network in which multiple modifiers converge on shared substrates such as APP and tau. This “modifier matrix” integrates signals across amyloid, tau, and inflammatory pathways to coordinate protein fate and cellular responses. We propose that AD pathology is driven not by changes in individual UBLs, but by an imbalance within the ubiquitin/UBL network that disrupts proteostasis. A key emerging theme is that different UBL pathways exert distinct and sometimes opposing effects on core AD processes. For example, SUMO2/3 signaling appears largely protective in tau pathology, whereas ISG15 promotes neuroinflammatory and proteostatic impairment, and NEDDylation shows strong context-dependent effects on both amyloid processing and synaptic function. This functional divergence suggests that AD arises from a shift in the balance between modifiers with competing roles. Defining how these pathways interact, rather than studying them in isolation, will be critical for understanding how the modifier matrix is rewired during disease progression.

Although individual UBLs offer significant potential for therapeutic intervention in AD pathology, several critical questions persist. Which changes in UBL modifier pathways act as true disease drivers versus adaptive responses in human AD? How do factors such as genetic variation, the presence of additional co-morbid pathologies, and systemic inflammation alter the modifier matrix? Finally, what are safe upper and lower thresholds for manipulating each UBL to avoid compromising important pathway function? It also remains unclear which UBL alterations arise earliest in preclinical AD and whether their roles shift during disease progression from adaptive to maladaptive. An additional open question is how UBL pathways are influenced by genetic forms of AD. Mutations in APP, PSEN1, and PSEN2 alter amyloid processing, cellular stress responses, and proteostasis, which may differentially engage UBL systems. It is conceivable that distinct genetic backgrounds bias the modifier landscape toward specific UBL pathways, thereby contributing to disease heterogeneity. Systematic comparisons of UBL signatures across familial and sporadic AD cases will be important to address this possibility.

Addressing these questions will require mechanistic studies, development of biomarker pipelines and therapeutic innovation. Priorities include dissecting UBL modifier pathways in human brain-relevant systems, coupled with advanced proteomics to define substrates, chain types, and hybrid topologies. These efforts should further clarify how the UBL-modified proteome is altered in AD and what functional consequences these changes have for proteostasis, neuroinflammation, and neuronal vulnerability. The functional consequences of UBL modulation should be evaluated in disease-relevant models including transgenic rodents, or stem-cell derived neuronal cultures or organoids to establish stage- and cell-type specific effects. Second, clinically deployable biomarker assays are needed to quantitatively measure specific UBL conjugates in accessible biospecimens, either as stand-alone indicators or integrated with established AD markers. Such tools will be essential for resolving driver versus compensatory roles and for monitoring therapeutic engagement. Third, therapeutic development should prioritize approaches that fine-tune rather than completely inhibit UBL pathways. Small molecules, antisense oligonucleotides/siRNAs, and biologics must be engineered to allow graded modulation, with cell-type specific delivery. Together, these efforts would shift the UBL modifier matrix from a descriptive feature of pathology to defined biological targets for AD diagnosis and intervention.

In summary, UBL pathways provide a unifying framework linking Aβ and tau to neuroinflammation, organelle dysfunction, impaired autophagy and synaptic decline. Their intrinsic druggability makes them attractive therapeutic targets. However, meaningful clinical translation will require precise modulation tailored to specific UBLs, substrates, cell types, brain regions and disease stages. Such precision is essential for efficacy while preserving the fundamental physiological function of each pathway. Future work integrating multi-UBL signatures, mechanistic insight, and targeted therapeutic strategies may enable a transition from descriptive mapping of UBL pathways to intervention in AD.

Abbreviations

Aβ

Amyloid-beta

AD

Alzheimer’s disease

APP

Amyloid precursor protein

APP-BP1

APP-binding protein 1

ATG8

Autophagy-related protein 8

BACE1

Beta-secretase 1

BAG3

BCL2-associated athanogene 3

CRL

Cullin-RING ligase

CSF

Cerebrospinal fluid

ER

Endoplasmic reticulum

GABARAP

Gamma-aminobutyric acid receptor-associated protein

HDAC6

Histone deacetylase 6

IFN

Interferon

ISG15

Interferon-stimulated gene 15

LC3

microtubule-associated protein 1 light chain 3

NAE

NEDD8-activating enzyme

NEDD8

Neural precursor cell expressed developmentally downregulated 8

PIAS

Protein inhibitor of activated STAT

PS1

presenilin-1

SAE

SUMO1 activating enzyme

SENP

Sentrin/SUMO-specific proteases

STAT

Signal transducer and activator of transcription

SUMO

small ubiquitin-like modifier

TRIM

Tripartite motif-containing

UBA

Ubiquitin like modifier activating enzyme

UBLs

Ubiquitin-like proteins

UFM1

Ubiquitin-fold modifier 1

UFSP

UFM1-specific protease

URM1

Ubiquitin-related modifier 1

USP

Ubiquitin-specific peptidase

Authors contributions

T.Y., J.V. and F.C.F. conceptualized, wrote, and revised the manuscript, and W.S. reviewed and revised the manuscript. All authors read and approved the final manuscript.

Funding

This work was supported by the Florida Department of Health Ed and Ethel Moore Alzheimer’s disease research program grant 22A07 (to F.C.F.). F.C.F is further supported by a Research Catalyst Award from the Mayo Clinic Center for Biomedical Discovery and the Office of Belonging and The Michael J. Fox Foundation for Parkinson’s Research. T.Y. is supported by a New Investigator Award from NACC and Alzheimer’s Association [NIAP25-1446587], a Florida Department of Health – Ed and Ethel Moore Alzheimer’s grant (25A11) and a fellowship from the Alzheimer’s Association [AARF-22–974258]. W.S. is supported in part by the National Institutes of Health (NIH)/National Institute of Neurological Disorders and Stroke (NINDS) [RF1NS085070, R01NS110085, and U54NS110435], National Institute on Aging (NIA) [R56AG062556], the Department of Defense Congressionally Directed Medical Research Programs (CDMRP) [W81XWH-17–1-0248], Mayo Foundation, Mayo Clinic Robert and Arlene Kogod Center on Aging, The Michael J. Fox Foundation for Parkinson’s Research, The Ted Nash Long Life Foundation, and the American Parkinson Disease Association (APDA) Center for Advanced Research at Mayo Clinic Jacksonville.

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.